Add OpenClaw skills, platform kit, and template docs

Made-with: Cursor

.claude/settings.local.json

@@ -0,0 +1,11 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(xargs -I {} bash -c \"echo '=== {} ===' && ls -la '{}'\")",
+      "Bash(python scripts/main.py version)",
+      "Bash(python scripts/main.py health)",
+      "Bash(python scripts/main.py key list)",
+      "Bash(python scripts/main.py generate kimi \"测试\")"
+    ]
+  }
+}

CLAUDE.md

@@ -0,0 +1,112 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## What This Repo Is
+
+**OpenClaw** is a monorepo of installable "Skills" (plugins) for the Claw agent hosting platform (深圳匠厂科技). Each subdirectory is an independent skill that ships as an encrypted ZIP to jc2009.com and gets installed by a desktop/gateway host.
+
+Each skill is independently versioned and can have its own git remote (`http://120.25.191.12:3000/admin/<slug>.git`). The root `.git` tracks only the `skill-template` scaffold.
+
+## Running Skills
+
+No build step — all skills are plain Python 3.10+. Run directly:
+
+```bash
+# Health / version check (every skill exposes these)
+python scripts/skill_main.py health
+python scripts/skill_main.py version
+
+# Account manager
+python account-manager/scripts/main.py list all
+python account-manager/scripts/main.py add "搜狐号" "13800138000"
+
+# Content manager
+python content-manager/scripts/main.py get <id>
+
+# API key vault
+python api-key-vault/scripts/vault.py get 17track
+
+# Logistics
+python logistics-tracker/scripts/main.py <tracking_number>
+
+# Sohu publish
+python sohu-publisher/scripts/main.py <account_id> <article_id>
+```
+
+## Releasing a Skill
+
+From the skill's directory:
+
+```powershell
+.\release.ps1            # tag + push
+.\release.ps1 -DryRun    # preview only
+.\release.ps1 -AutoCommit -CommitMessage "Release v1.0.5"
+```
+
+This delegates to `jiangchang-platform-kit/tools/release.ps1`, creates a semver git tag, and triggers the GitHub Actions workflow that encrypts scripts via PyArmor, packs a ZIP, and syncs metadata to jc2009.com.
+
+## Architecture
+
+### Skill Structure
+
+Every skill directory contains:
+- `SKILL.md` — YAML frontmatter (name, version, slug, category, dependencies, allowed-tools) + Markdown body with usage triggers and env requirements. **Version here must stay in sync with the git tag.**
+- `scripts/` — Python entry points
+- `release.ps1` — delegates to platform kit
+
+### Environment Contract (`CLAW_*`)
+
+Skills locate user data via env vars:
+
+| Variable | Purpose | Dev default |
+|---|---|---|
+| `CLAW_DATA_ROOT` | User data root | `D:\claw-data` or `~/.claw-data` |
+| `CLAW_USER_ID` | Workspace ID | `_anon` |
+| `CLAW_SKILLS_ROOT` | Sibling skills dir | Parent of skill dir |
+
+Legacy aliases (`JIANGCHANG_DATA_ROOT`, etc.) are still detected; prefer `CLAW_*` in new code.
+
+Data path convention: `{CLAW_DATA_ROOT}/{CLAW_USER_ID}/{slug}/` (SQLite DBs, cache, temp).
+
+### Skill Types
+
+| Type | Pattern | Examples |
+|---|---|---|
+| A | Stateless tool | logistics-tracker |
+| B | Local persistent (SQLite) | account-manager, content-manager, api-key-vault |
+| C | Orchestrator (calls sibling skills via subprocess) | — |
+| D | Hybrid (B + C) | sohu-publisher |
+
+Type C/D skills declare `dependencies.required` in SKILL.md and call siblings like:
+```python
+subprocess.run(["python", f"{skills_root}/account-manager/scripts/main.py", ...])
+```
+
+### Shared Platform Kit (`jiangchang-platform-kit/`)
+
+- `sdk/jiangchang_skill_core` — `EntitlementClient`, `enforce_entitlement` (license checks)
+- `.github/workflows/reusable-release-skill.yaml` — reusable CI called by every skill's release workflow
+- `tools/release.ps1` — shared release script
+
+### CLI Conventions
+
+- Subcommands: `health` (offline quick check, exits 0/1), `version` (JSON output), custom actions
+- Success prefix: `OK` or plain output
+- Error prefix: `ERROR:` (e.g., `ERROR:REQUIRE_LOGIN`)
+- Windows GBK fix: wrap stdout/stderr with `io.TextIOWrapper(..., encoding='utf-8')`
+
+## Key Docs
+
+| File | What it covers |
+|---|---|
+| `skill-template/README.md` | Onboarding, anti-patterns |
+| `skill-template/docs/RUNTIME.md` | Full env var contract and fallback behavior |
+| `skill-template/docs/SKILL_TYPES.md` | Type A–D classification checklist |
+| `skill-template/docs/PORTABILITY.md` | Cross-platform encoding, path differences |
+| `skill-template/docs/LOGGING.md` | `TimedRotatingFileHandler` setup |
+
+## Known Dev-Mode Issues
+
+- `account-manager/scripts/main.py` has `_ACCOUNT_MANAGER_CLI_LOCAL_DEV = True` hardcoded with a fixed data path (`D:\jiangchang-data`) and user ID — disable before releasing.
+- `toutiao-publisher` is a placeholder stub; publish logic is not yet implemented.

account-manager/.github/workflows/release_skill.yaml

@@ -0,0 +1,11 @@
+name: 技能自动化发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

account-manager/SKILL.md

@@ -0,0 +1,53 @@
+---
+name: 账号管理
+description: 多平台多账号管理。管理各平台账号与Chrome Profile的对应关系，供publisher类Skill调用获取账号信息。
+version: 1.0.4
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: account-manager
+    emoji: "👤"
+    category: "通用"
+allowed-tools:
+  - bash
+---
+
+# 账号管理
+
+## 使用时机
+
+当用户发送以下内容的时候触发本Skill：
+- 说"初始化账号登录"、"登录搜狐账号"、"帮我登录账号"
+- 说"列出账号"、"查看所有账号"、"有哪些账号"
+- 说"添加账号"、"删除账号"、"获取账号信息"
+
+## 执行步骤
+
+### 添加账号（平台用中文名；须填合法中国大陆 11 位手机号；同平台下同手机号不可重复；ID 自增，名称如「搜狐1号」）
+```bash
+python3 {baseDir}/scripts/main.py add 搜狐号 189xxxxxxxx
+python3 {baseDir}/scripts/main.py add 知乎 13800138000
+```
+支持的平台称呼示例：搜狐号、头条号、知乎、微信公众号、Kimi、DeepSeek、豆包、通义千问、文心一言、腾讯元宝（亦支持英文键如 sohu、toutiao）。
+
+### 仅打开浏览器核对是否已登录（不写数据库）
+```bash
+python3 {baseDir}/scripts/main.py open <id>
+```
+
+### 登录并自动检测、写入数据库
+```bash
+python3 {baseDir}/scripts/main.py login <id>
+```
+
+### 删除账号（同时删库里的记录与 profile 用户数据目录）
+```bash
+python3 {baseDir}/scripts/main.py delete id <id>
+python3 {baseDir}/scripts/main.py delete platform <平台>
+python3 {baseDir}/scripts/main.py delete platform <平台> <手机号>
+```
+
+### 列出某平台所有账号（platform 可为中文名、all 或 全部）
+```bash
+python3 {baseDir}/scripts/main.py list all
+python3 {baseDir}/scripts/main.py list 搜狐号

account-manager/release.ps1

@@ -0,0 +1,23 @@
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+
+if (-not (Test-Path $sharedScript)) {
+  throw "Shared release script not found: $sharedScript"
+}
+
+& $sharedScript @PSBoundParameters
+exit $LASTEXITCODE

api-key-vault/.github/workflows/release_skill.yaml

@@ -0,0 +1,11 @@
+name: 技能自动化发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

api-key-vault/SKILL.md

@@ -0,0 +1,54 @@
+---
+name: API Key管理
+description: API Key统一管理工具。用于存储、读取、更新、删除各种第三方平台的API Key。当需要获取任何平台的API Key时调用本Skill。
+version: 1.0.0
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: api-key-vault
+    emoji: "🔐"
+    category: "通用"
+allowed-tools:
+  - bash
+---
+
+# API Key 管理vault
+
+## 用途
+
+统一管理所有第三方平台的API Key，供其他Skill调用。
+Key存储在本地 `.env` 文件中，不上传任何服务器。
+
+## 使用方式
+
+### 读取一个Key
+```bash
+python3 {baseDir}/scripts/vault.py get 17track
+```
+
+### 写入/更新一个Key
+```bash
+python3 {baseDir}/scripts/vault.py set 17track YOUR_API_KEY_HERE
+```
+
+### 列出所有已存储的Key名称
+```bash
+python3 {baseDir}/scripts/vault.py list
+```
+
+### 删除一个Key
+```bash
+python3 {baseDir}/scripts/vault.py delete 17track
+```
+
+## 返回格式
+
+- get成功：直接返回Key的值，无多余内容
+- get失败：返回 `ERROR:KEY_NOT_FOUND`
+- set/delete/list：返回操作结果说明
+
+## 注意事项
+
+- Key名称统一用小写+连字符，例如 `17track`、`ups-api`、`fedex-oauth`
+- `.env` 文件位于 `api-key-vault/` 根目录，不要手动编辑格式
+- 任何Skill需要Key时，调用本Skill的get命令获取，不要硬编码

api-key-vault/release.ps1

@@ -0,0 +1,220 @@
+<#
+.SYNOPSIS
+  One-command release script for skill repos.
+
+.DESCRIPTION
+  - Optional auto-commit
+  - Push current branch
+  - Auto-increment semantic tag (vX.Y.Z)
+  - Create & push tag
+  - Fail fast on unsafe states
+
+.EXAMPLES
+  # Safe mode (recommended): requires clean working tree
+  .\release.ps1
+
+  # Auto commit tracked/untracked changes then release
+  .\release.ps1 -AutoCommit -CommitMessage "chore: update skill config"
+
+  # Dry run (show what would happen)
+  .\release.ps1 -DryRun
+
+  # Custom tag prefix
+  .\release.ps1 -Prefix "v" -Message "正式发布"
+
+.NOTES
+  Requires: git, PowerShell 5+
+#>
+
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+if (Test-Path $sharedScript) {
+  & $sharedScript @PSBoundParameters
+  exit $LASTEXITCODE
+}
+
+function Invoke-Git {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  Write-Host ">> git $Args" -ForegroundColor DarkGray
+  & cmd /c "git $Args"
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+}
+
+function Get-GitOutput {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  $output = & cmd /c "git $Args" 2>$null
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+  return @($output)
+}
+
+function Test-Repo {
+  & git rev-parse --is-inside-work-tree *> $null
+  return ($LASTEXITCODE -eq 0)
+}
+
+function Get-CurrentBranch {
+  $b = (Get-GitOutput "branch --show-current" | Select-Object -First 1).Trim()
+  return $b
+}
+
+function Get-StatusPorcelain {
+  $lines = @(Get-GitOutput "status --porcelain")
+  return $lines
+}
+
+function Parse-SemVerTag {
+  param(
+    [string]$Tag,
+    [string]$TagPrefix
+  )
+  $escaped = [regex]::Escape($TagPrefix)
+  $m = [regex]::Match($Tag, "^${escaped}(\d+)\.(\d+)\.(\d+)$")
+  if (-not $m.Success) { return $null }
+
+  return [pscustomobject]@{
+    Raw   = $Tag
+    Major = [int]$m.Groups[1].Value
+    Minor = [int]$m.Groups[2].Value
+    Patch = [int]$m.Groups[3].Value
+  }
+}
+
+function Get-NextTag {
+  param([string]$TagPrefix)
+
+  $tags = Get-GitOutput "tag --list"
+  $parsed = @()
+
+  foreach ($t in $tags) {
+    $t = $t.Trim()
+    if (-not $t) { continue }
+    $obj = Parse-SemVerTag -Tag $t -TagPrefix $TagPrefix
+    if ($null -ne $obj) { $parsed += $obj }
+  }
+
+  if ($parsed.Count -eq 0) {
+    return "${TagPrefix}1.0.1"
+  }
+
+  $latest = $parsed | Sort-Object Major, Minor, Patch | Select-Object -Last 1
+  return "$TagPrefix$($latest.Major).$($latest.Minor).$([int]$latest.Patch + 1)"
+}
+
+function Ensure-CleanOrAutoCommit {
+  param(
+    [switch]$DoAutoCommit,
+    [switch]$NeedClean,
+    [switch]$IsDryRun,
+    [string]$Msg
+  )
+
+  $status = @(Get-StatusPorcelain)
+  if ($status.Length -eq 0) { return }
+
+  if ($NeedClean) {
+    Write-Host "Working tree is not clean and -RequireClean is enabled." -ForegroundColor Yellow
+    & git status --short
+    throw "Abort: dirty working tree."
+  }
+
+  # 默认一键发布：有改动就自动提交；也可用 -AutoCommit 显式开启
+  $commitMsg = $Msg
+  if ([string]::IsNullOrWhiteSpace($commitMsg)) {
+    $commitMsg = "chore: auto release commit ($(Get-Date -Format 'yyyy-MM-dd HH:mm:ss'))"
+  }
+
+  if (-not $DoAutoCommit) {
+    Write-Host "Detected uncommitted changes, auto-committing before release..." -ForegroundColor Yellow
+  }
+
+  if ($IsDryRun) {
+    Write-Host "[DryRun] Would run: git add -A" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git commit -m `"$commitMsg`"" -ForegroundColor Yellow
+    return
+  }
+
+  Invoke-Git "add -A"
+  Invoke-Git "commit -m `"$commitMsg`""
+}
+
+try {
+  Write-Host "=== Release Script Start ===" -ForegroundColor Cyan
+
+  if (-not (Test-Repo)) {
+    throw "Current directory is not a git repository."
+  }
+
+  $branch = Get-CurrentBranch
+  if ([string]::IsNullOrWhiteSpace($branch)) {
+    throw "Unable to determine current branch."
+  }
+
+  if ($branch -notin @("main", "master")) {
+    throw "Current branch is '$branch'. Release is only allowed from main/master."
+  }
+
+  Invoke-Git "fetch --tags --prune origin"
+
+  Ensure-CleanOrAutoCommit -DoAutoCommit:$AutoCommit -NeedClean:$RequireClean -IsDryRun:$DryRun -Msg $CommitMessage
+
+  $upstream = (& git rev-parse --abbrev-ref --symbolic-full-name "@{u}" 2>$null)
+  $hasUpstream = ($LASTEXITCODE -eq 0)
+
+  if ($DryRun) {
+    if ($hasUpstream) {
+      Write-Host "[DryRun] Would run: git push" -ForegroundColor Yellow
+    } else {
+      Write-Host "[DryRun] Would run: git push -u origin $branch" -ForegroundColor Yellow
+    }
+  } else {
+    if ($hasUpstream) {
+      Invoke-Git "push"
+    } else {
+      Invoke-Git "push -u origin $branch"
+    }
+  }
+
+  $nextTag = Get-NextTag -TagPrefix $Prefix
+  Write-Host "Next tag: $nextTag" -ForegroundColor Green
+
+  $existing = @(Get-GitOutput "tag --list `"$nextTag`"")
+  if ($existing.Length -gt 0) {
+    throw "Tag already exists: $nextTag"
+  }
+
+  if ($DryRun) {
+    Write-Host "[DryRun] Would run: git tag -a $nextTag -m `"$Message`"" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git push origin $nextTag" -ForegroundColor Yellow
+    Write-Host "=== DryRun Complete ===" -ForegroundColor Cyan
+    exit 0
+  }
+
+  Invoke-Git "tag -a $nextTag -m `"$Message`""
+  Invoke-Git "push origin $nextTag"
+
+  Write-Host "Release success: $nextTag" -ForegroundColor Green
+  Write-Host "=== Release Script Done ===" -ForegroundColor Cyan
+  exit 0
+}
+catch {
+  Write-Host "Release failed: $($_.Exception.Message)" -ForegroundColor Red
+  exit 1
+}

api-key-vault/scripts/vault.py

@@ -0,0 +1,105 @@
+import sys
+import os
+
+# .env文件路径：固定在api-key-vault根目录
+BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+ENV_FILE = os.path.join(BASE_DIR, ".env")
+
+
+def load_keys():
+    """读取.env文件，返回dict"""
+    keys = {}
+    if not os.path.exists(ENV_FILE):
+        return keys
+    with open(ENV_FILE, "r", encoding="utf-8") as f:
+        for line in f:
+            line = line.strip()
+            if not line or line.startswith("#"):
+                continue
+            if "=" in line:
+                name, _, value = line.partition("=")
+                keys[name.strip()] = value.strip()
+    return keys
+
+
+def save_keys(keys):
+    """把dict写回.env文件"""
+    with open(ENV_FILE, "w", encoding="utf-8") as f:
+        f.write("# API Key Vault - 自动生成，请勿手动乱改\n")
+        for name, value in keys.items():
+            f.write(f"{name}={value}\n")
+
+
+def cmd_get(name):
+    keys = load_keys()
+    if name not in keys:
+        print("ERROR:KEY_NOT_FOUND")
+    else:
+        print(keys[name])
+
+
+def cmd_set(name, value):
+    keys = load_keys()
+    keys[name] = value
+    save_keys(keys)
+    print(f"✅ 已保存：{name}")
+
+
+def cmd_list():
+    keys = load_keys()
+    if not keys:
+        print("暂无已存储的Key")
+        return
+    print("已存储的Key列表：")
+    for name in keys:
+        value = keys[name]
+        # 只显示前4位和后4位，中间用*遮挡
+        masked = value[:4] + "****" + value[-4:] if len(value) > 8 else "****"
+        print(f"  · {name} = {masked}")
+
+
+def cmd_delete(name):
+    keys = load_keys()
+    if name not in keys:
+        print(f"❌ 未找到：{name}")
+    else:
+        del keys[name]
+        save_keys(keys)
+        print(f"🗑 已删除：{name}")
+
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        print("用法：")
+        print("  python vault.py get <key名>")
+        print("  python vault.py set <key名> <key值>")
+        print("  python vault.py list")
+        print("  python vault.py delete <key名>")
+        sys.exit(1)
+
+    command = sys.argv[1].lower()
+
+    if command == "get":
+        if len(sys.argv) < 3:
+            print("错误：get命令需要提供key名")
+            sys.exit(1)
+        cmd_get(sys.argv[2])
+
+    elif command == "set":
+        if len(sys.argv) < 4:
+            print("错误：set命令需要提供key名和key值")
+            sys.exit(1)
+        cmd_set(sys.argv[2], sys.argv[3])
+
+    elif command == "list":
+        cmd_list()
+
+    elif command == "delete":
+        if len(sys.argv) < 3:
+            print("错误：delete命令需要提供key名")
+            sys.exit(1)
+        cmd_delete(sys.argv[2])
+
+    else:
+        print(f"错误：未知命令 {command}")
+        sys.exit(1)

content-manager/.github/workflows/release_skill.yaml

@@ -0,0 +1,11 @@
+name: 技能自动化发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

content-manager/SKILL.md

@@ -0,0 +1,82 @@
+---
+name: 内容管理
+description: 文章、图片、视频分表分目录管理；文章正文在库内，图/视频文件在数据目录、库内仅存相对路径。代码分层：db 仓储 / services 业务 / cli 入口。
+version: 2.0.0
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: content-manager
+    emoji: "📝"
+    category: "通用"
+    dependencies:
+      required:
+        - llm-manager
+      auto_install: false
+allowed-tools:
+  - bash
+---
+
+# 内容管理器（文章 / 图片 / 视频）
+
+## 代码结构（MVC 分层）
+
+- `scripts/main.py`：仅入口与环境准备（`sys.path`、Windows UTF-8）。
+- `content_manager/cli/`：参数解析与分发（Controller）。
+- `content_manager/services/`：业务编排（调用仓储、llm-manager、文件复制）。
+- `content_manager/db/`：SQLite 连接、建表迁移、按表划分的 **repository**（只做 SQL，不含业务规则）。
+- `content_manager/config.py`：数据根路径、技能数据目录。
+- `content_manager/constants.py`：提示词种子、平台别名等常量。
+
+## 数据与表结构
+
+- 库路径：`{JIANGCHANG_DATA_ROOT}/{JIANGCHANG_USER_ID}/content-manager/content-manager.db`
+- **文章** 表 `articles`：`title`、`body`、`content_html`、`status`、`source`、`llm_target`、`account_id`、`error_msg`、`extra_json`、时间戳等（正文在库内）。
+- **图片** 表 `images`：仅存 **`file_path`（相对技能数据目录）** 及 `title`、`status`、`source` 等元数据；二进制在 `{…}/content-manager/images/<id>/`。
+- **视频** 表 `videos`：同上，文件在 `{…}/content-manager/videos/<id>/`；可选 `duration_ms`。
+- 提示词相关表：`prompt_templates`、`prompt_template_usage`（供文章 `generate`）。
+
+从旧版「TEXT 主键」`articles` 库启动时会自动迁移到新结构。
+
+## 常用命令
+
+将 `{baseDir}` 换为技能根目录。一级子命令为 **`article` / `image` / `video`**。
+
+### 文章
+
+```bash
+python {baseDir}/scripts/main.py article list
+python {baseDir}/scripts/main.py article get <id>
+python {baseDir}/scripts/main.py article add --title "标题" --body "正文"
+python {baseDir}/scripts/main.py article add --title "标题" --body-file D:\path\article.md
+python {baseDir}/scripts/main.py article import-json D:\path\articles.json
+python {baseDir}/scripts/main.py article generate 豆包 搜狐号 "RPA降本增效"
+python {baseDir}/scripts/main.py article prompt-list 搜狐号 --limit 20
+python {baseDir}/scripts/main.py article delete <id>
+python {baseDir}/scripts/main.py article feedback <id> published <account_id>
+python {baseDir}/scripts/main.py article save <id或占位> <title> <单行正文>
+```
+
+`article get` 输出 JSON：`id`、`title`、`content`、`content_html`、`status` 等。
+
+### 图片 / 视频（库内只存路径）
+
+```bash
+python {baseDir}/scripts/main.py image add --file D:\a.png [--title "说明"]
+python {baseDir}/scripts/main.py image list
+python {baseDir}/scripts/main.py image get <id>
+python {baseDir}/scripts/main.py image delete <id>
+python {baseDir}/scripts/main.py image feedback <id> published <account_id>
+
+python {baseDir}/scripts/main.py video add --file D:\a.mp4 [--title "说明"] [--duration-ms 120000]
+python {baseDir}/scripts/main.py video list
+python {baseDir}/scripts/main.py video get <id>
+python {baseDir}/scripts/main.py video delete <id>
+python {baseDir}/scripts/main.py video feedback <id> failed <account_id> "原因"
+```
+
+`image get` / `video get` 的 JSON 含 `file_path`（相对）、`absolute_path`（解析后绝对路径）。
+
+## 环境变量
+
+- `JIANGCHANG_DATA_ROOT`、`JIANGCHANG_USER_ID`：与 account-manager 一致。
+- `llm-manager` 依赖其自身环境与账号/API Key 配置。

content-manager/content_manager/__init__.py

@@ -0,0 +1 @@
+# content-manager 技能：文章 / 图片 / 视频 分层包

content-manager/content_manager/cli/__init__.py

@@ -0,0 +1 @@
+# CLI：解析参数并调用 services

content-manager/content_manager/cli/app.py

@@ -0,0 +1,261 @@
+"""CLI 入口：argparse 装配与分发（Controller）。"""
+
+from __future__ import annotations
+
+import argparse
+import os
+import sys
+from typing import List, Optional
+
+from content_manager.services import article_service, image_service, video_service
+from content_manager.services.article_service import resolve_publish_platform
+from content_manager.util.argparse_zh import ZhArgumentParser
+
+
+def _print_root_usage_zh() -> None:
+    print(
+        """内容管理：请指定资源类型子命令。
+
+  python main.py article list
+  python main.py article get 1
+  python main.py article add --title "标题" --body-file 文章.md
+  python main.py article generate 豆包 搜狐号 RPA降本增效
+  python main.py image add --file D:\\\\a.png [--title "说明"]
+  python main.py video add --file D:\\\\a.mp4 [--title "说明"] [--duration-ms 120000]
+
+查看完整说明：python main.py -h"""
+    )
+
+
+def _handle_article_add(args: argparse.Namespace) -> None:
+    if args.body_file:
+        fp = os.path.abspath(args.body_file)
+        try:
+            with open(fp, encoding="utf-8") as f:
+                body = f.read()
+        except OSError as e:
+            print(f"❌ 无法读取正文文件：{fp}\n原因：{e}")
+            sys.exit(1)
+    else:
+        body = args.body or ""
+    article_service.cmd_add(args.title, body, source="manual")
+
+
+def _handle_article_import(args: argparse.Namespace) -> None:
+    article_service.cmd_import_json(args.path)
+
+
+def _handle_article_generate(args: argparse.Namespace) -> None:
+    raw_parts = [str(x).strip() for x in (args.generate_args or []) if str(x).strip()]
+    if not raw_parts:
+        print("❌ 缺少主题或关键词。")
+        print("示例：python main.py article generate 豆包 搜狐号 RPA降本增效")
+        sys.exit(1)
+    platform_guess = resolve_publish_platform(raw_parts[0])
+    if platform_guess and len(raw_parts) == 1:
+        print("❌ 缺少主题或关键词。")
+        sys.exit(1)
+    if platform_guess and len(raw_parts) >= 2:
+        publish_platform = platform_guess
+        topic = " ".join(raw_parts[1:]).strip()
+    else:
+        publish_platform = "common"
+        topic = " ".join(raw_parts).strip()
+    if not topic:
+        print("❌ 主题或关键词不能为空。")
+        sys.exit(1)
+    article_service.cmd_generate(
+        args.llm_target,
+        topic,
+        publish_platform=publish_platform,
+        title=getattr(args, "title", None),
+    )
+
+
+def _handle_article_feedback(args: argparse.Namespace) -> None:
+    article_service.cmd_feedback(args.article_id, args.status, args.account_id, args.error_msg)
+
+
+def _handle_article_save_legacy(args: argparse.Namespace) -> None:
+    article_service.cmd_save(args.legacy_id, args.legacy_title, args.legacy_content)
+
+
+def _handle_image_add(args: argparse.Namespace) -> None:
+    image_service.cmd_add(args.file, title=getattr(args, "title", None))
+
+
+def _handle_image_feedback(args: argparse.Namespace) -> None:
+    image_service.cmd_feedback(args.image_id, args.status, args.account_id, args.error_msg)
+
+
+def _handle_video_add(args: argparse.Namespace) -> None:
+    video_service.cmd_add(
+        args.file,
+        title=getattr(args, "title", None),
+        duration_ms=getattr(args, "duration_ms", None),
+    )
+
+
+def _handle_video_feedback(args: argparse.Namespace) -> None:
+    video_service.cmd_feedback(args.video_id, args.status, args.account_id, args.error_msg)
+
+
+def build_parser() -> ZhArgumentParser:
+    fmt = argparse.RawDescriptionHelpFormatter
+    p = ZhArgumentParser(
+        prog="main.py",
+        description="内容管理：文章（正文在库内）与图片/视频（文件在数据目录，库内仅存路径）。",
+        epilog="示例见各子命令 -h；一级分组：article / image / video",
+        formatter_class=fmt,
+    )
+    sub = p.add_subparsers(
+        dest="resource",
+        required=True,
+        metavar="资源类型",
+        help="article 文章 | image 图片 | video 视频",
+        parser_class=ZhArgumentParser,
+    )
+
+    # ----- article -----
+    art = sub.add_parser("article", help="文章：正文与元数据在 SQLite", formatter_class=fmt)
+    art_sub = art.add_subparsers(
+        dest="article_cmd",
+        required=True,
+        metavar="子命令",
+        parser_class=ZhArgumentParser,
+    )
+
+    sp = art_sub.add_parser("list", help="列出文章")
+    sp.add_argument("--limit", type=int, default=10)
+    sp.add_argument("--max-chars", type=int, default=50)
+    sp.set_defaults(handler=lambda a: article_service.cmd_list(limit=a.limit, max_chars=a.max_chars))
+
+    sp = art_sub.add_parser("get", help="按 id 输出 JSON")
+    sp.add_argument("article_id", metavar="文章id")
+    sp.set_defaults(handler=lambda a: article_service.cmd_get(a.article_id))
+
+    sp = art_sub.add_parser(
+        "add",
+        help="新增文章",
+        formatter_class=fmt,
+        epilog="示例：python main.py article add --title \"标题\" --body \"正文\"",
+    )
+    sp.add_argument("--title", required=True)
+    g = sp.add_mutually_exclusive_group(required=True)
+    g.add_argument("--body-file", metavar="路径")
+    g.add_argument("--body", metavar="正文")
+    sp.set_defaults(handler=_handle_article_add)
+
+    sp = art_sub.add_parser("import-json", help="从 JSON 批量导入")
+    sp.add_argument("path", metavar="JSON路径")
+    sp.set_defaults(handler=_handle_article_import)
+
+    sp = art_sub.add_parser("generate", help="调用 llm-manager 生成并入库", formatter_class=fmt)
+    sp.add_argument("llm_target", metavar="大模型目标")
+    sp.add_argument("generate_args", nargs="+", metavar="生成参数")
+    sp.add_argument("--title", metavar="标题", default=None)
+    sp.set_defaults(handler=_handle_article_generate)
+
+    sp = art_sub.add_parser("prompt-list", help="查看提示词模板")
+    sp.add_argument("platform", nargs="?", default=None, metavar="发布平台")
+    sp.add_argument("--limit", type=int, default=30)
+    sp.set_defaults(handler=lambda a: article_service.cmd_prompt_list(a.platform, a.limit))
+
+    sp = art_sub.add_parser("delete", help="删除文章")
+    sp.add_argument("article_id", metavar="文章id")
+    sp.set_defaults(handler=lambda a: article_service.cmd_delete(a.article_id))
+
+    sp = art_sub.add_parser("feedback", help="回写发布状态", formatter_class=fmt)
+    sp.add_argument("article_id", metavar="文章id")
+    sp.add_argument("status", metavar="状态")
+    sp.add_argument("account_id", nargs="?", default=None, metavar="账号")
+    sp.add_argument("error_msg", nargs="?", default=None, metavar="错误说明")
+    sp.set_defaults(handler=_handle_article_feedback)
+
+    sp = art_sub.add_parser("save", help="旧版单行正文保存", formatter_class=fmt)
+    sp.add_argument("legacy_id", metavar="id")
+    sp.add_argument("legacy_title", metavar="标题")
+    sp.add_argument("legacy_content", metavar="正文一行")
+    sp.set_defaults(handler=_handle_article_save_legacy)
+
+    # ----- image -----
+    img = sub.add_parser("image", help="图片：文件在数据目录，images 表存相对路径", formatter_class=fmt)
+    img_sub = img.add_subparsers(
+        dest="image_cmd",
+        required=True,
+        metavar="子命令",
+        parser_class=ZhArgumentParser,
+    )
+
+    sp = img_sub.add_parser("list", help="列出图片")
+    sp.add_argument("--limit", type=int, default=20)
+    sp.add_argument("--max-chars", type=int, default=80)
+    sp.set_defaults(handler=lambda a: image_service.cmd_list(limit=a.limit, max_chars=a.max_chars))
+
+    sp = img_sub.add_parser("get", help="按 id 输出 JSON（含 absolute_path）")
+    sp.add_argument("image_id", metavar="图片id")
+    sp.set_defaults(handler=lambda a: image_service.cmd_get(a.image_id))
+
+    sp = img_sub.add_parser("add", help="从本地文件复制入库", formatter_class=fmt)
+    sp.add_argument("--file", required=True, metavar="文件", help="源图片路径")
+    sp.add_argument("--title", default=None, metavar="标题", help="可选说明")
+    sp.set_defaults(handler=_handle_image_add)
+
+    sp = img_sub.add_parser("delete", help="删除记录与磁盘目录")
+    sp.add_argument("image_id", metavar="图片id")
+    sp.set_defaults(handler=lambda a: image_service.cmd_delete(a.image_id))
+
+    sp = img_sub.add_parser("feedback", help="回写状态", formatter_class=fmt)
+    sp.add_argument("image_id", metavar="图片id")
+    sp.add_argument("status", metavar="状态")
+    sp.add_argument("account_id", nargs="?", default=None, metavar="账号")
+    sp.add_argument("error_msg", nargs="?", default=None, metavar="错误说明")
+    sp.set_defaults(handler=_handle_image_feedback)
+
+    # ----- video -----
+    vid = sub.add_parser("video", help="视频：文件在数据目录，videos 表存相对路径", formatter_class=fmt)
+    vid_sub = vid.add_subparsers(
+        dest="video_cmd",
+        required=True,
+        metavar="子命令",
+        parser_class=ZhArgumentParser,
+    )
+
+    sp = vid_sub.add_parser("list", help="列出视频")
+    sp.add_argument("--limit", type=int, default=20)
+    sp.add_argument("--max-chars", type=int, default=80)
+    sp.set_defaults(handler=lambda a: video_service.cmd_list(limit=a.limit, max_chars=a.max_chars))
+
+    sp = vid_sub.add_parser("get", help="按 id 输出 JSON")
+    sp.add_argument("video_id", metavar="视频id")
+    sp.set_defaults(handler=lambda a: video_service.cmd_get(a.video_id))
+
+    sp = vid_sub.add_parser("add", help="从本地文件复制入库", formatter_class=fmt)
+    sp.add_argument("--file", required=True, metavar="文件")
+    sp.add_argument("--title", default=None, metavar="标题")
+    sp.add_argument("--duration-ms", type=int, default=None, metavar="毫秒", help="可选时长")
+    sp.set_defaults(handler=_handle_video_add)
+
+    sp = vid_sub.add_parser("delete", help="删除记录与磁盘目录")
+    sp.add_argument("video_id", metavar="视频id")
+    sp.set_defaults(handler=lambda a: video_service.cmd_delete(a.video_id))
+
+    sp = vid_sub.add_parser("feedback", help="回写状态", formatter_class=fmt)
+    sp.add_argument("video_id", metavar="视频id")
+    sp.add_argument("status", metavar="状态")
+    sp.add_argument("account_id", nargs="?", default=None, metavar="账号")
+    sp.add_argument("error_msg", nargs="?", default=None, metavar="错误说明")
+    sp.set_defaults(handler=_handle_video_feedback)
+
+    return p
+
+
+def main(argv: Optional[List[str]] = None) -> int:
+    argv = argv if argv is not None else sys.argv[1:]
+    if not argv:
+        _print_root_usage_zh()
+        return 1
+    parser = build_parser()
+    args = parser.parse_args(argv)
+    args.handler(args)
+    return 0

content-manager/content_manager/config.py

@@ -0,0 +1,47 @@
+"""路径与环境：与 account-manager 一致的数据根目录。"""
+
+from __future__ import annotations
+
+import os
+import sys
+from typing import Optional
+
+SKILL_SLUG = "content-manager"
+
+
+def get_skill_root() -> str:
+    # content_manager/config.py -> 技能根 content-manager/
+    return os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+
+
+def get_openclaw_root() -> str:
+    return os.path.dirname(get_skill_root())
+
+
+def get_data_root() -> str:
+    env = (os.getenv("JIANGCHANG_DATA_ROOT") or "").strip()
+    if env:
+        return env
+    if sys.platform == "win32":
+        return r"D:\jiangchang-data"
+    return os.path.join(os.path.expanduser("~"), ".jiangchang-data")
+
+
+def get_user_id() -> str:
+    return (os.getenv("JIANGCHANG_USER_ID") or "").strip() or "_anon"
+
+
+def get_skill_data_dir() -> str:
+    path = os.path.join(get_data_root(), get_user_id(), SKILL_SLUG)
+    os.makedirs(path, exist_ok=True)
+    return path
+
+
+def get_db_path() -> str:
+    return os.path.join(get_skill_data_dir(), "content-manager.db")
+
+
+def resolve_stored_path(relative_file_path: str) -> str:
+    """库内相对路径 -> 绝对路径。"""
+    rel = (relative_file_path or "").strip().replace("\\", "/").lstrip("/")
+    return os.path.normpath(os.path.join(get_skill_data_dir(), rel))

content-manager/content_manager/constants.py

@@ -0,0 +1,103 @@
+"""与业务相关的常量（提示词种子、平台别名、CLI 提示文案）。"""
+
+from __future__ import annotations
+
+from typing import Dict, List, Set
+
+CLI_REQUIRED_ZH = {
+    "cmd": "一级子命令：article 文章 | image 图片 | video 视频",
+    "子命令": "二级子命令，用 -h 查看该分组下的命令",
+    "llm_target": "大模型目标：写平台名（如 豆包、DeepSeek、Kimi）或 account-manager 里已登录账号的纯数字 id",
+    "大模型目标": "大模型目标：写平台名（如 豆包、DeepSeek、Kimi）或 account-manager 里已登录账号的纯数字 id",
+    "生成参数": "生成参数：格式是「模型 [发布平台] 主题/关键词」，例如：python main.py article generate 豆包 搜狐号 RPA降本增效",
+    "主题": "主题或关键词：至少填写一项，用于自动套用提示词模板",
+    "--title": "标题：写 --title \"文章标题\"；add / generate / image add / video add 会用到",
+    "标题": "标题：写 --title \"文章标题\"",
+    "--body": "正文：写 --body \"短文\"；与 --body-file 二选一",
+    "正文": "正文：写 --body \"短文\"；与 --body-file 二选一",
+    "--body-file": "正文文件：写 --body-file 后再写 UTF-8 文件路径；与 --body 二选一",
+    "路径": "正文文件：写 --body-file 后再写 UTF-8 文件路径；与 --body 二选一",
+    "--file": "本地文件路径：图片或视频源文件",
+    "文件": "本地文件路径：图片或视频源文件",
+    "path": "JSON 文件路径：写在 import-json 后面，例如 D:\\\\data\\\\articles.json",
+    "JSON路径": "JSON 文件路径：写在 import-json 后面，例如 D:\\\\data\\\\articles.json",
+    "article_id": "文章编号：整数 id，可先执行 article list 看最左一列",
+    "文章id": "文章编号：整数 id，可先执行 article list 看最左一列",
+    "image_id": "图片编号：整数 id，可先执行 image list",
+    "图片id": "图片编号：整数 id，可先执行 image list",
+    "video_id": "视频编号：整数 id，可先执行 video list",
+    "视频id": "视频编号：整数 id，可先执行 video list",
+    "legacy_id": "id：若是已有文章的数字 id 则更新；否则新建一篇",
+    "id": "id：若是已有文章的数字 id 则更新；否则新建一篇",
+    "legacy_title": "标题",
+    "legacy_content": "正文（整段须在一行内，不要换行）",
+    "正文一行": "正文（整段须在一行内，不要换行）",
+    "状态": "状态：例如 published（已发布）或 failed（失败）",
+    "账号": "账号标识：可选，给发布记录用",
+    "错误说明": "错误说明：可选，发布失败时写上原因",
+}
+
+PUBLISH_PLATFORM_ALIASES: Dict[str, Set[str]] = {
+    "common": {"common", "通用", "默认", "general", "all"},
+    "sohu": {"sohu", "搜狐", "搜狐号"},
+    "toutiao": {"toutiao", "头条", "头条号", "今日头条"},
+    "wechat": {"wechat", "weixin", "wx", "公众号", "微信公众号", "微信"},
+}
+
+PUBLISH_PLATFORM_CN = {
+    "common": "通用",
+    "sohu": "搜狐号",
+    "toutiao": "头条号",
+    "wechat": "微信公众号",
+}
+
+PROMPT_TEMPLATE_SEEDS: Dict[str, List[str]] = {
+    "common": [
+        "请围绕主题“{topic}”写一篇结构完整、可直接发布的新媒体文章，输出纯正文，不要解释。",
+        "请以“{topic}”为核心，写一篇适合中文互联网平台发布的文章，语言自然、观点清晰、可读性强。",
+        "围绕“{topic}”写一篇实用向内容，要求有标题、导语、分点展开和结语，整体逻辑清楚。",
+        "请写一篇关于“{topic}”的科普文章，面向普通读者，避免术语堆砌，语气专业但易懂。",
+        "请从痛点、原因、方法、案例四个部分展开，写一篇主题为“{topic}”的原创内容。",
+        "围绕“{topic}”写一篇信息密度高但不枯燥的文章，要求段落清晰、句子简洁。",
+        "请就“{topic}”写一篇观点型文章，先给结论，再给依据和建议，最后总结。",
+        "请生成一篇主题为“{topic}”的内容，适合移动端阅读，段落不宜过长，便于快速浏览。",
+        "围绕“{topic}”撰写一篇可发布文章，避免空话套话，优先给出可执行建议。",
+        "请以“{topic}”为题写文，要求开头抓人、中段有干货、结尾有行动建议。",
+    ],
+    "sohu": [
+        "你在为搜狐号写稿。请围绕“{topic}”写一篇原创文章，风格稳重、信息扎实，适合搜狐号读者阅读。",
+        "请按搜狐号内容风格，围绕“{topic}”写一篇逻辑清晰、观点明确的文章，输出纯正文。",
+        "面向搜狐号发布场景，生成主题“{topic}”文章，要求有吸引力标题和清晰分段。",
+        "请写一篇搜狐号可发布稿件，主题“{topic}”，强调实用价值与可读性。",
+        "围绕“{topic}”写搜狐号文章：先引出问题，再给分析，最后给建议。",
+        "请生成一篇适配搜狐号用户阅读习惯的文章，主题是“{topic}”，语言自然且有深度。",
+        "请写一篇“{topic}”主题稿，适合搜狐号发布，避免口水话，突出真实信息和案例。",
+        "为搜狐号生成“{topic}”文章，结构为：导语-正文三段-总结，输出可直接发布内容。",
+        "请围绕“{topic}”写一篇搜狐号文章，强调观点清晰、段落层次分明、结尾有启发。",
+        "围绕“{topic}”产出搜狐号稿件，内容原创、连贯、可读，避免模板化表达。",
+    ],
+    "toutiao": [
+        "请按头条号风格围绕“{topic}”写文章，开头3句要抓人，正文信息密度高。",
+        "请写一篇头条号可发布内容，主题“{topic}”，要求标题感强、节奏快、观点明确。",
+        "围绕“{topic}”写头条稿件，语言更口语化、易传播，适当加入场景化表达。",
+        "请生成“{topic}”头条文章：开头抛问题，中段拆解，结尾给结论。",
+        "为头条号创作“{topic}”文章，注重读者停留与完读，段落短、信息集中。",
+        "请写一篇“{topic}”主题头条文，强调实用技巧和可执行方法。",
+        "围绕“{topic}”生成头条风格内容，避免空泛，突出细节与案例。",
+        "请按头条读者偏好写“{topic}”文章，语气直接，观点鲜明，结尾有行动建议。",
+        "请围绕“{topic}”写头条号稿件，确保逻辑清楚、表达简洁、节奏紧凑。",
+        "写一篇适合头条号发布的“{topic}”文章，要求易懂、好读、可传播。",
+    ],
+    "wechat": [
+        "请按公众号长文风格围绕“{topic}”写稿，语气克制、叙述完整、可深度阅读。",
+        "请写一篇适合公众号发布的“{topic}”文章，包含引言、分节标题和总结。",
+        "围绕“{topic}”写公众号文章，强调逻辑深度与观点完整性，输出纯正文。",
+        "请创作“{topic}”公众号稿件，要求有故事化开头、干货正文、结尾金句。",
+        "请按公众号读者习惯，写一篇主题“{topic}”的内容，表达自然、层次清晰。",
+        "生成一篇“{topic}”微信公众号文章，强调洞察与方法论，避免碎片化表达。",
+        "请围绕“{topic}”写公众号稿，风格专业可信，段落清楚且有小标题。",
+        "请写一篇“{topic}”公众号内容，结构为：问题提出-原因分析-解决建议-结语。",
+        "请生成“{topic}”公众号文章，注重阅读体验，段落与节奏适合移动端。",
+        "围绕“{topic}”撰写可直接发公众号的文章，要求原创、完整、可读。",
+    ],
+}

content-manager/content_manager/db/__init__.py

@@ -0,0 +1,3 @@
+from content_manager.db.connection import get_conn, init_db
+
+__all__ = ["get_conn", "init_db"]

content-manager/content_manager/db/articles_repository.py

@@ -0,0 +1,122 @@
+"""articles 表：仅负责 SQL 读写，不含业务规则。"""
+
+from __future__ import annotations
+
+import sqlite3
+from typing import Any, List, Optional, Tuple
+
+
+def insert_article(
+    conn: sqlite3.Connection,
+    title: str,
+    body: str,
+    content_html: Optional[str],
+    status: str,
+    source: str,
+    account_id: Optional[str],
+    error_msg: Optional[str],
+    llm_target: Optional[str],
+    extra_json: Optional[str],
+    created_at: int,
+    updated_at: int,
+) -> int:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        INSERT INTO articles (
+            title, body, content_html, status, source, account_id, error_msg, llm_target, extra_json,
+            created_at, updated_at
+        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+        """,
+        (
+            title,
+            body,
+            content_html,
+            status,
+            source,
+            account_id,
+            error_msg,
+            llm_target,
+            extra_json,
+            created_at,
+            updated_at,
+        ),
+    )
+    return int(cur.lastrowid)
+
+
+def update_article_body(
+    conn: sqlite3.Connection,
+    article_id: int,
+    title: str,
+    body: str,
+    updated_at: int,
+) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        UPDATE articles SET title = ?, body = ?, updated_at = ?
+        WHERE id = ?
+        """,
+        (title, body, updated_at, article_id),
+    )
+
+
+def fetch_by_id(conn: sqlite3.Connection, article_id: int) -> Optional[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, title, body, content_html, status, source, account_id, error_msg,
+               llm_target, extra_json, created_at, updated_at
+        FROM articles WHERE id = ?
+        """,
+        (article_id,),
+    )
+    return cur.fetchone()
+
+
+def exists_id(conn: sqlite3.Connection, article_id: int) -> bool:
+    cur = conn.cursor()
+    cur.execute("SELECT id FROM articles WHERE id = ?", (article_id,))
+    return cur.fetchone() is not None
+
+
+def list_recent(conn: sqlite3.Connection, limit: int) -> List[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT
+            id, title, body, content_html,
+            status, source, account_id, error_msg, llm_target, extra_json,
+            created_at, updated_at
+        FROM articles ORDER BY updated_at DESC, id DESC
+        LIMIT ?
+        """,
+        (int(limit),),
+    )
+    return list(cur.fetchall())
+
+
+def delete_by_id(conn: sqlite3.Connection, article_id: int) -> int:
+    cur = conn.cursor()
+    cur.execute("DELETE FROM articles WHERE id = ?", (article_id,))
+    return int(cur.rowcount)
+
+
+def update_feedback(
+    conn: sqlite3.Connection,
+    article_id: int,
+    status: str,
+    account_id: Optional[str],
+    error_msg: Optional[str],
+    updated_at: int,
+) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        UPDATE articles
+        SET status = ?, account_id = ?, error_msg = ?, updated_at = ?
+        WHERE id = ?
+        """,
+        (status, account_id, error_msg, updated_at, article_id),
+    )

content-manager/content_manager/db/connection.py

@@ -0,0 +1,113 @@
+"""数据库连接与初始化（建表、文章旧库迁移、提示词种子）。"""
+
+from __future__ import annotations
+
+import sqlite3
+from typing import TYPE_CHECKING
+
+from content_manager.config import get_db_path
+from content_manager.db.schema import (
+    ARTICLES_TABLE_SQL,
+    IMAGES_TABLE_SQL,
+    PROMPT_TEMPLATE_USAGE_TABLE_SQL,
+    PROMPT_TEMPLATES_TABLE_SQL,
+    VIDEOS_TABLE_SQL,
+)
+from content_manager.util.timeutil import now_unix, parse_ts_to_unix
+
+if TYPE_CHECKING:
+    pass
+
+
+def get_conn() -> sqlite3.Connection:
+    return sqlite3.connect(get_db_path())
+
+
+def _is_legacy_articles_table(cur: sqlite3.Cursor) -> bool:
+    cur.execute("PRAGMA table_info(articles)")
+    rows = cur.fetchall()
+    if not rows:
+        return False
+    for _cid, name, ctype, _nn, _d, _pk in rows:
+        if name == "id" and ctype and ctype.upper() == "INTEGER":
+            return False
+    cur.execute("SELECT sql FROM sqlite_master WHERE type='table' AND name='articles'")
+    row = cur.fetchone()
+    if row and row[0] and "TEXT" in row[0] and "id" in row[0]:
+        return True
+    return False
+
+
+def _migrate_legacy_articles(conn: sqlite3.Connection) -> None:
+    cur = conn.cursor()
+    cur.executescript(
+        """
+        CREATE TABLE _articles_migrated (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            title TEXT NOT NULL,
+            body TEXT NOT NULL,
+            content_html TEXT,
+            status TEXT NOT NULL DEFAULT 'draft',
+            source TEXT NOT NULL DEFAULT 'manual',
+            account_id TEXT,
+            error_msg TEXT,
+            llm_target TEXT,
+            extra_json TEXT,
+            created_at INTEGER NOT NULL,
+            updated_at INTEGER NOT NULL
+        );
+        """
+    )
+    cur.execute(
+        "SELECT id, title, content, content_html, status, account_id, error_msg, created_at, updated_at FROM articles"
+    )
+    ts = now_unix()
+    for row in cur.fetchall():
+        _oid, title, content, content_html, status, account_id, error_msg, cat, uat = row
+        body = content or ""
+        ch = content_html if content_html else None
+        cts = parse_ts_to_unix(cat) or ts
+        uts = parse_ts_to_unix(uat) or ts
+        cur.execute(
+            """
+            INSERT INTO _articles_migrated (
+                title, body, content_html, status, source, account_id, error_msg,
+                created_at, updated_at
+            ) VALUES (?, ?, ?, ?, 'import', ?, ?, ?, ?)
+            """,
+            (
+                title or "",
+                body,
+                ch,
+                (status or "draft"),
+                account_id,
+                error_msg,
+                cts,
+                uts,
+            ),
+        )
+    cur.execute("DROP TABLE articles")
+    cur.execute("ALTER TABLE _articles_migrated RENAME TO articles")
+    conn.commit()
+
+
+def init_db() -> None:
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='articles'")
+        if cur.fetchone():
+            if _is_legacy_articles_table(cur):
+                _migrate_legacy_articles(conn)
+        else:
+            cur.executescript(ARTICLES_TABLE_SQL)
+        cur.executescript(IMAGES_TABLE_SQL)
+        cur.executescript(VIDEOS_TABLE_SQL)
+        cur.executescript(PROMPT_TEMPLATES_TABLE_SQL)
+        cur.executescript(PROMPT_TEMPLATE_USAGE_TABLE_SQL)
+        from content_manager.db.prompts_repository import seed_prompt_templates_if_empty
+
+        seed_prompt_templates_if_empty(conn.cursor())
+        conn.commit()
+    finally:
+        conn.close()

content-manager/content_manager/db/images_repository.py

@@ -0,0 +1,88 @@
+"""images 表：仅保存文件相对路径等元数据。"""
+
+from __future__ import annotations
+
+import sqlite3
+from typing import Any, List, Optional, Tuple
+
+
+def insert_row(
+    conn: sqlite3.Connection,
+    file_path: str,
+    title: Optional[str],
+    status: str,
+    source: str,
+    account_id: Optional[str],
+    error_msg: Optional[str],
+    extra_json: Optional[str],
+    created_at: int,
+    updated_at: int,
+) -> int:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        INSERT INTO images (
+            file_path, title, status, source, account_id, error_msg, extra_json, created_at, updated_at
+        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+        """,
+        (file_path, title, status, source, account_id, error_msg, extra_json, created_at, updated_at),
+    )
+    return int(cur.lastrowid)
+
+
+def update_file_path(conn: sqlite3.Connection, image_id: int, file_path: str, updated_at: int) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        "UPDATE images SET file_path = ?, updated_at = ? WHERE id = ?",
+        (file_path, updated_at, image_id),
+    )
+
+
+def fetch_by_id(conn: sqlite3.Connection, image_id: int) -> Optional[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, file_path, title, status, source, account_id, error_msg, extra_json, created_at, updated_at
+        FROM images WHERE id = ?
+        """,
+        (image_id,),
+    )
+    return cur.fetchone()
+
+
+def list_recent(conn: sqlite3.Connection, limit: int) -> List[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, file_path, title, status, source, account_id, error_msg, extra_json, created_at, updated_at
+        FROM images ORDER BY updated_at DESC, id DESC
+        LIMIT ?
+        """,
+        (int(limit),),
+    )
+    return list(cur.fetchall())
+
+
+def delete_by_id(conn: sqlite3.Connection, image_id: int) -> int:
+    cur = conn.cursor()
+    cur.execute("DELETE FROM images WHERE id = ?", (image_id,))
+    return int(cur.rowcount)
+
+
+def update_feedback(
+    conn: sqlite3.Connection,
+    image_id: int,
+    status: str,
+    account_id: Optional[str],
+    error_msg: Optional[str],
+    updated_at: int,
+) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        UPDATE images
+        SET status = ?, account_id = ?, error_msg = ?, updated_at = ?
+        WHERE id = ?
+        """,
+        (status, account_id, error_msg, updated_at, image_id),
+    )

content-manager/content_manager/db/prompts_repository.py

@@ -0,0 +1,114 @@
+"""提示词模板：表内数据访问与种子。"""
+
+from __future__ import annotations
+
+import random
+import sqlite3
+from typing import Any, Dict, List, Optional, Tuple
+
+from content_manager.constants import PROMPT_TEMPLATE_SEEDS, PUBLISH_PLATFORM_CN
+from content_manager.util.timeutil import now_unix
+
+
+def seed_prompt_templates_if_empty(cur: sqlite3.Cursor) -> None:
+    ts = now_unix()
+    for platform, templates in PROMPT_TEMPLATE_SEEDS.items():
+        cur.execute("SELECT COUNT(*) FROM prompt_templates WHERE platform = ?", (platform,))
+        count = int(cur.fetchone()[0] or 0)
+        if count > 0:
+            continue
+        for idx, tpl in enumerate(templates, start=1):
+            cur.execute(
+                """
+                INSERT INTO prompt_templates (platform, name, template_text, is_active, created_at, updated_at)
+                VALUES (?, ?, ?, 1, ?, ?)
+                """,
+                (platform, f"{PUBLISH_PLATFORM_CN.get(platform, platform)}模板{idx}", tpl, ts, ts),
+            )
+
+
+def count_by_platform(cur: sqlite3.Cursor, platform: str) -> int:
+    cur.execute("SELECT COUNT(*) FROM prompt_templates WHERE platform = ?", (platform,))
+    return int(cur.fetchone()[0] or 0)
+
+
+def fetch_active_templates(conn: sqlite3.Connection, platform: str) -> List[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, platform, name, template_text
+        FROM prompt_templates
+        WHERE platform = ? AND is_active = 1
+        ORDER BY id ASC
+        """,
+        (platform,),
+    )
+    return list(cur.fetchall())
+
+
+def fetch_common_fallback(conn: sqlite3.Connection) -> List[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, platform, name, template_text
+        FROM prompt_templates
+        WHERE platform = 'common' AND is_active = 1
+        ORDER BY id ASC
+        """
+    )
+    return list(cur.fetchall())
+
+
+def pick_random_template(rows: List[Tuple[Any, ...]]) -> Optional[Dict[str, Any]]:
+    if not rows:
+        return None
+    rid, p, name, text = random.choice(rows)
+    return {"id": int(rid), "platform": p, "name": name, "template_text": text}
+
+
+def insert_usage(
+    conn: sqlite3.Connection,
+    template_id: int,
+    llm_target: str,
+    platform: str,
+    topic: str,
+    article_id: Optional[int],
+) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        INSERT INTO prompt_template_usage (template_id, llm_target, platform, topic, article_id, created_at)
+        VALUES (?, ?, ?, ?, ?, ?)
+        """,
+        (template_id, llm_target, platform, topic, article_id, now_unix()),
+    )
+
+
+def list_templates(
+    conn: sqlite3.Connection,
+    platform: Optional[str],
+    limit: int,
+) -> List[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    if platform:
+        cur.execute(
+            """
+            SELECT id, platform, name, is_active, updated_at
+            FROM prompt_templates
+            WHERE platform = ?
+            ORDER BY id DESC
+            LIMIT ?
+            """,
+            (platform, int(limit)),
+        )
+    else:
+        cur.execute(
+            """
+            SELECT id, platform, name, is_active, updated_at
+            FROM prompt_templates
+            ORDER BY id DESC
+            LIMIT ?
+            """,
+            (int(limit),),
+        )
+    return list(cur.fetchall())

content-manager/content_manager/db/schema.py

@@ -0,0 +1,73 @@
+"""建表 SQL（不含迁移逻辑）。"""
+
+ARTICLES_TABLE_SQL = """
+CREATE TABLE IF NOT EXISTS articles (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    title TEXT NOT NULL,
+    body TEXT NOT NULL,
+    content_html TEXT,
+    status TEXT NOT NULL DEFAULT 'draft',
+    source TEXT NOT NULL DEFAULT 'manual',
+    account_id TEXT,
+    error_msg TEXT,
+    llm_target TEXT,
+    extra_json TEXT,
+    created_at INTEGER NOT NULL,
+    updated_at INTEGER NOT NULL
+);
+"""
+
+IMAGES_TABLE_SQL = """
+CREATE TABLE IF NOT EXISTS images (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    file_path TEXT NOT NULL,
+    title TEXT,
+    status TEXT NOT NULL DEFAULT 'draft',
+    source TEXT NOT NULL DEFAULT 'manual',
+    account_id TEXT,
+    error_msg TEXT,
+    extra_json TEXT,
+    created_at INTEGER NOT NULL,
+    updated_at INTEGER NOT NULL
+);
+"""
+
+VIDEOS_TABLE_SQL = """
+CREATE TABLE IF NOT EXISTS videos (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    file_path TEXT NOT NULL,
+    title TEXT,
+    duration_ms INTEGER,
+    status TEXT NOT NULL DEFAULT 'draft',
+    source TEXT NOT NULL DEFAULT 'manual',
+    account_id TEXT,
+    error_msg TEXT,
+    extra_json TEXT,
+    created_at INTEGER NOT NULL,
+    updated_at INTEGER NOT NULL
+);
+"""
+
+PROMPT_TEMPLATES_TABLE_SQL = """
+CREATE TABLE IF NOT EXISTS prompt_templates (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    platform TEXT NOT NULL,
+    name TEXT NOT NULL,
+    template_text TEXT NOT NULL,
+    is_active INTEGER NOT NULL DEFAULT 1,
+    created_at INTEGER NOT NULL,
+    updated_at INTEGER NOT NULL
+);
+"""
+
+PROMPT_TEMPLATE_USAGE_TABLE_SQL = """
+CREATE TABLE IF NOT EXISTS prompt_template_usage (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    template_id INTEGER NOT NULL,
+    llm_target TEXT NOT NULL,
+    platform TEXT NOT NULL,
+    topic TEXT NOT NULL,
+    article_id INTEGER,
+    created_at INTEGER NOT NULL
+);
+"""

content-manager/content_manager/db/videos_repository.py

@@ -0,0 +1,100 @@
+"""videos 表：仅保存文件相对路径等元数据。"""
+
+from __future__ import annotations
+
+import sqlite3
+from typing import Any, List, Optional, Tuple
+
+
+def insert_row(
+    conn: sqlite3.Connection,
+    file_path: str,
+    title: Optional[str],
+    duration_ms: Optional[int],
+    status: str,
+    source: str,
+    account_id: Optional[str],
+    error_msg: Optional[str],
+    extra_json: Optional[str],
+    created_at: int,
+    updated_at: int,
+) -> int:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        INSERT INTO videos (
+            file_path, title, duration_ms, status, source, account_id, error_msg, extra_json, created_at, updated_at
+        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+        """,
+        (
+            file_path,
+            title,
+            duration_ms,
+            status,
+            source,
+            account_id,
+            error_msg,
+            extra_json,
+            created_at,
+            updated_at,
+        ),
+    )
+    return int(cur.lastrowid)
+
+
+def update_file_path(conn: sqlite3.Connection, video_id: int, file_path: str, updated_at: int) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        "UPDATE videos SET file_path = ?, updated_at = ? WHERE id = ?",
+        (file_path, updated_at, video_id),
+    )
+
+
+def fetch_by_id(conn: sqlite3.Connection, video_id: int) -> Optional[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, file_path, title, duration_ms, status, source, account_id, error_msg, extra_json, created_at, updated_at
+        FROM videos WHERE id = ?
+        """,
+        (video_id,),
+    )
+    return cur.fetchone()
+
+
+def list_recent(conn: sqlite3.Connection, limit: int) -> List[Tuple[Any, ...]]:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        SELECT id, file_path, title, duration_ms, status, source, account_id, error_msg, extra_json, created_at, updated_at
+        FROM videos ORDER BY updated_at DESC, id DESC
+        LIMIT ?
+        """,
+        (int(limit),),
+    )
+    return list(cur.fetchall())
+
+
+def delete_by_id(conn: sqlite3.Connection, video_id: int) -> int:
+    cur = conn.cursor()
+    cur.execute("DELETE FROM videos WHERE id = ?", (video_id,))
+    return int(cur.rowcount)
+
+
+def update_feedback(
+    conn: sqlite3.Connection,
+    video_id: int,
+    status: str,
+    account_id: Optional[str],
+    error_msg: Optional[str],
+    updated_at: int,
+) -> None:
+    cur = conn.cursor()
+    cur.execute(
+        """
+        UPDATE videos
+        SET status = ?, account_id = ?, error_msg = ?, updated_at = ?
+        WHERE id = ?
+        """,
+        (status, account_id, error_msg, updated_at, video_id),
+    )

content-manager/content_manager/services/__init__.py

@@ -0,0 +1 @@
+# 业务逻辑层（调用 db 仓储，不含 argparse）

content-manager/content_manager/services/article_service.py

@@ -0,0 +1,431 @@
+"""文章：业务规则与编排（调用仓储 + llm-manager）。"""
+
+from __future__ import annotations
+
+import json
+import os
+import re
+import subprocess
+import sys
+from typing import Any, Dict, Optional
+
+from content_manager.config import get_openclaw_root
+from content_manager.constants import PUBLISH_PLATFORM_CN, PUBLISH_PLATFORM_ALIASES
+from content_manager.db import articles_repository as ar
+from content_manager.db import prompts_repository as pr
+from content_manager.db.connection import get_conn, init_db
+from content_manager.util.timeutil import now_unix, unix_to_iso
+
+
+def _row_to_public_dict(row: tuple) -> Dict[str, Any]:
+    rid, title, body, content_html, status, source, account_id, error_msg, llm_target, extra_json, cat, uat = row
+    d: Dict[str, Any] = {
+        "id": int(rid),
+        "title": title,
+        "content": body,
+        "content_html": content_html if content_html else body,
+        "status": status or "draft",
+        "source": source or "manual",
+        "account_id": account_id,
+        "error_msg": error_msg,
+        "llm_target": llm_target,
+        "created_at": unix_to_iso(cat),
+        "updated_at": unix_to_iso(uat),
+    }
+    if extra_json:
+        try:
+            ex = json.loads(extra_json)
+            if isinstance(ex, dict):
+                d["extra"] = ex
+        except json.JSONDecodeError:
+            pass
+    return d
+
+
+def resolve_publish_platform(raw: Optional[str]) -> Optional[str]:
+    s = (raw or "").strip().lower()
+    if not s:
+        return "common"
+    for key, aliases in PUBLISH_PLATFORM_ALIASES.items():
+        if s in {a.lower() for a in aliases}:
+            return key
+    return None
+
+
+def _choose_prompt_template(platform: str) -> Optional[Dict[str, Any]]:
+    init_db()
+    conn = get_conn()
+    try:
+        rows = pr.fetch_active_templates(conn, platform)
+        if not rows and platform != "common":
+            rows = pr.fetch_common_fallback(conn)
+    finally:
+        conn.close()
+    return pr.pick_random_template(rows)
+
+
+def _build_prompt_from_template(template_text: str, topic: str, platform: str) -> str:
+    platform_name = PUBLISH_PLATFORM_CN.get(platform, "通用")
+    rendered = (
+        template_text.replace("{topic}", topic).replace("{platform}", platform).replace("{platform_name}", platform_name)
+    )
+    return rendered.strip()
+
+
+def cmd_add(title: str, body: str, source: str = "manual", llm_target: Optional[str] = None) -> None:
+    init_db()
+    title = (title or "").strip() or "未命名"
+    body = body or ""
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        new_id = ar.insert_article(
+            conn,
+            title=title,
+            body=body,
+            content_html=None,
+            status="draft",
+            source=source,
+            account_id=None,
+            error_msg=None,
+            llm_target=llm_target,
+            extra_json=None,
+            created_at=ts,
+            updated_at=ts,
+        )
+        conn.commit()
+    finally:
+        conn.close()
+    print(f"✅ 已新增文章 id={new_id} | {title}")
+
+
+def cmd_import_json(path: str) -> None:
+    init_db()
+    path = os.path.abspath(path.strip())
+    if not os.path.isfile(path):
+        print(f"❌ 找不到文件：{path}\n请检查路径是否正确、文件是否存在。")
+        sys.exit(1)
+    with open(path, encoding="utf-8") as f:
+        raw = json.load(f)
+    if isinstance(raw, dict) and "articles" in raw:
+        items = raw["articles"]
+    elif isinstance(raw, list):
+        items = raw
+    else:
+        print(
+            "❌ JSON 格式不对。\n"
+            "正确格式二选一：① 文件里是数组 [ {\"title\":\"…\",\"body\":\"…\"}, … ]\n"
+            "② 或对象 {\"articles\": [ … ] }，数组里每项至少要有正文（body 或 content）。"
+        )
+        sys.exit(1)
+    if not items:
+        print("❌ JSON 里没有可导入的文章条目（数组为空）。")
+        sys.exit(1)
+    n = 0
+    for i, item in enumerate(items):
+        if not isinstance(item, dict):
+            print(f"❌ 第 {i + 1} 条不是 JSON 对象（应为 {{ \"title\":…, \"body\":… }}）。")
+            sys.exit(1)
+        title = (item.get("title") or item.get("标题") or "").strip()
+        body = item.get("body") or item.get("content") or item.get("正文") or ""
+        if isinstance(body, dict):
+            print(f"❌ 第 {i + 1} 条的 body/content 必须是字符串，不能是别的类型。")
+            sys.exit(1)
+        body = str(body)
+        if not title and not body.strip():
+            continue
+        if not title:
+            title = f"导入-{i + 1}"
+        cmd_add(title, body, source="import")
+        n += 1
+    print(f"✅ 批量导入完成，共写入 {n} 篇")
+
+
+def _parse_llm_stdout(stdout: str) -> str:
+    if "===LLM_START===" in stdout and "===LLM_END===" in stdout:
+        chunk = stdout.split("===LLM_START===", 1)[1]
+        chunk = chunk.split("===LLM_END===", 1)[0]
+        return chunk.strip()
+    return (stdout or "").strip()
+
+
+def _default_title_from_body(body: str) -> str:
+    for line in body.splitlines():
+        t = line.strip()
+        if t:
+            return t[:120] if len(t) > 120 else t
+    return f"文稿-{now_unix()}"
+
+
+def cmd_generate(
+    llm_target: str,
+    topic: str,
+    publish_platform: str = "common",
+    title: Optional[str] = None,
+) -> None:
+    llm_target = (llm_target or "").strip()
+    topic = (topic or "").strip()
+    publish_platform = (publish_platform or "common").strip().lower()
+    if not llm_target or not topic:
+        print(
+            "❌ 生成参数不完整。\n"
+            "请使用：python main.py article generate <模型> [发布平台] <主题或关键词>\n"
+            "示例：python main.py article generate 豆包 搜狐号 RPA降本增效"
+        )
+        sys.exit(1)
+    template = _choose_prompt_template(publish_platform)
+    if not template:
+        print("❌ 提示词模板库为空，请先补充模板后再执行 generate。")
+        sys.exit(1)
+    prompt = _build_prompt_from_template(template["template_text"], topic, publish_platform)
+    script = os.path.join(get_openclaw_root(), "llm-manager", "scripts", "main.py")
+    if not os.path.isfile(script):
+        print(
+            f"❌ 找不到大模型脚本：{script}\n"
+            "请确认 llm-manager 与 content-manager 在同一上级目录（OpenClaw）下。"
+        )
+        sys.exit(1)
+    proc = subprocess.run(
+        [sys.executable, script, "generate", llm_target, prompt],
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+    )
+    out = (proc.stdout or "") + "\n" + (proc.stderr or "")
+    std = proc.stdout or ""
+    has_markers = "===LLM_START===" in std and "===LLM_END===" in std
+    if (proc.returncode != 0 and not has_markers) or (
+        proc.returncode == 0 and not has_markers and re.search(r"(?m)^ERROR:", std)
+    ):
+        print(
+            (out.strip() or f"大模型进程退出码 {proc.returncode}")
+            + "\n❌ 生成失败：请根据上面说明处理（常见：先在「模型管理」添加并登录该平台账号，或配置 API Key）。"
+        )
+        sys.exit(1)
+    body = _parse_llm_stdout(proc.stdout or out)
+    if not body:
+        print(
+            "❌ 没有从大模型输出里取到正文。\n"
+            "正常情况输出里应包含 ===LLM_START=== 与 ===LLM_END===；请重试或查看 llm-manager 是否正常打印。"
+        )
+        sys.exit(1)
+    body = body.strip()
+    if body.startswith("ERROR:"):
+        print(out.strip())
+        print(f"\n❌ 生成失败，未写入数据库。\n{body}")
+        sys.exit(1)
+    final_title = (title or "").strip() or _default_title_from_body(body)
+    extra_payload = {
+        "generate_meta": {
+            "mode": "template",
+            "topic": topic,
+            "platform": publish_platform,
+            "platform_cn": PUBLISH_PLATFORM_CN.get(publish_platform, publish_platform),
+            "template_id": template["id"],
+            "template_name": template["name"],
+        }
+    }
+    init_db()
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        new_id = ar.insert_article(
+            conn,
+            title=final_title,
+            body=body,
+            content_html=None,
+            status="draft",
+            source="llm",
+            account_id=None,
+            error_msg=None,
+            llm_target=llm_target,
+            extra_json=json.dumps(extra_payload, ensure_ascii=False),
+            created_at=ts,
+            updated_at=ts,
+        )
+        pr.insert_usage(conn, int(template["id"]), llm_target, publish_platform, topic, int(new_id))
+        conn.commit()
+    finally:
+        conn.close()
+    print(
+        f"✅ 已写入 LLM 文稿 id={new_id} | {final_title}\n"
+        f"   模板：{template['name']} (id={template['id']}) | 平台：{PUBLISH_PLATFORM_CN.get(publish_platform, publish_platform)} | 主题：{topic}"
+    )
+
+
+def cmd_prompt_list(platform: Optional[str] = None, limit: int = 30) -> None:
+    init_db()
+    if limit <= 0:
+        limit = 30
+    key = resolve_publish_platform(platform) if platform else None
+    if platform and not key:
+        print(f"❌ 不支持的平台：{platform}")
+        print("支持：通用 / 搜狐号 / 头条号 / 公众号")
+        sys.exit(1)
+    conn = get_conn()
+    try:
+        rows = pr.list_templates(conn, key, limit)
+    finally:
+        conn.close()
+    if not rows:
+        print("暂无提示词模板")
+        return
+    sep_line = "_" * 39
+    for idx, (rid, p, name, active, uat) in enumerate(rows):
+        print(f"id：{rid}")
+        print(f"platform：{p}")
+        print(f"platform_cn：{PUBLISH_PLATFORM_CN.get(p, p)}")
+        print(f"name：{name}")
+        print(f"is_active：{int(active)}")
+        print(f"updated_at：{unix_to_iso(uat) or ''}")
+        if idx != len(rows) - 1:
+            print(sep_line)
+            print()
+
+
+def cmd_save(article_id: str, title: str, content: str) -> None:
+    init_db()
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        if article_id.isdigit():
+            aid = int(article_id)
+            if ar.exists_id(conn, aid):
+                ar.update_article_body(conn, aid, title, content, ts)
+                conn.commit()
+                print(f"✅ 已更新 id={aid} | {title}")
+                return
+        new_id = ar.insert_article(
+            conn,
+            title=title,
+            body=content,
+            content_html=None,
+            status="draft",
+            source="manual",
+            account_id=None,
+            error_msg=None,
+            llm_target=None,
+            extra_json=None,
+            created_at=ts,
+            updated_at=ts,
+        )
+        conn.commit()
+        print(f"✅ 已新建 id={new_id} | {title}")
+    finally:
+        conn.close()
+
+
+def cmd_get(article_id: str) -> None:
+    init_db()
+    if not str(article_id).strip().isdigit():
+        print("❌ 文章 id 必须是纯数字（整数）。请先 article list 查看最左一列编号。")
+        sys.exit(1)
+    aid = int(article_id)
+    conn = get_conn()
+    try:
+        row = ar.fetch_by_id(conn, aid)
+    finally:
+        conn.close()
+    if not row:
+        print("❌ 没有这篇文章：该 id 在库里不存在。请先执行 article list 核对编号。")
+        sys.exit(1)
+    print(json.dumps(_row_to_public_dict(row), ensure_ascii=False))
+
+
+def cmd_list(limit: int = 10, max_chars: int = 50) -> None:
+    init_db()
+    conn = get_conn()
+    try:
+        rows = ar.list_recent(conn, limit)
+    finally:
+        conn.close()
+    if not rows:
+        print("暂无文章")
+        return
+
+    def maybe_truncate(text: str) -> str:
+        if not text:
+            return ""
+        if len(text) > max_chars:
+            return text[:max_chars] + "..."
+        return text
+
+    sep_line = "_" * 39
+    for idx, r in enumerate(rows):
+        (
+            rid,
+            title,
+            body,
+            content_html,
+            status,
+            source,
+            account_id,
+            error_msg,
+            llm_target,
+            extra_json,
+            created_at,
+            updated_at,
+        ) = r
+        content = content_html if content_html else (body or "")
+
+        print(f"id：{rid}")
+        print(f"title：{title or ''}")
+        print("body：")
+        print(maybe_truncate(body or ""))
+        print("content：")
+        print(maybe_truncate(content or ""))
+        print(f"status：{status or ''}")
+        print(f"source：{source or ''}")
+        print(f"account_id：{account_id or ''}")
+        print(f"error_msg：{error_msg or ''}")
+        print(f"llm_target：{llm_target or ''}")
+        print(f"extra_json：{extra_json or ''}")
+        print(f"created_at：{unix_to_iso(created_at) or ''}")
+        print(f"updated_at：{unix_to_iso(updated_at) or ''}")
+
+        if idx != len(rows) - 1:
+            print(sep_line)
+            print()
+
+
+def cmd_delete(article_id: str) -> None:
+    init_db()
+    if not str(article_id).strip().isdigit():
+        print("❌ 文章 id 必须是纯数字。请先 article list 查看。")
+        sys.exit(1)
+    aid = int(article_id)
+    conn = get_conn()
+    try:
+        n = ar.delete_by_id(conn, aid)
+        if n == 0:
+            print("❌ 没有 id 为 {} 的文章，无法删除。".format(aid))
+            sys.exit(1)
+        conn.commit()
+    finally:
+        conn.close()
+    print(f"✅ 已删除 id={aid}")
+
+
+def cmd_feedback(
+    article_id: str,
+    status: str,
+    account_id: Optional[str] = None,
+    error_msg: Optional[str] = None,
+) -> None:
+    init_db()
+    if not str(article_id).strip().isdigit():
+        print("❌ 文章 id 必须是纯数字。")
+        sys.exit(1)
+    aid = int(article_id)
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        if not ar.exists_id(conn, aid):
+            print("❌ 没有 id 为 {} 的文章，无法回写状态。".format(aid))
+            sys.exit(1)
+        ar.update_feedback(conn, aid, status, account_id, error_msg, ts)
+        conn.commit()
+    finally:
+        conn.close()
+    print("✅ 状态已更新")

content-manager/content_manager/services/file_store.py

@@ -0,0 +1,50 @@
+"""媒体文件落盘：相对技能数据目录的路径约定。"""
+
+from __future__ import annotations
+
+import os
+import shutil
+from typing import Tuple
+
+
+def media_subdir(kind: str, media_id: int) -> str:
+    """kind: images | videos"""
+    return f"{kind}/{media_id}"
+
+
+def original_basename(src_path: str) -> str:
+    ext = os.path.splitext(src_path)[1]
+    return f"original{ext if ext else ''}"
+
+
+def copy_into_skill_data(
+    skill_data_dir: str,
+    kind: str,
+    media_id: int,
+    src_path: str,
+) -> Tuple[str, str]:
+    """
+    将源文件复制到 {skill_data_dir}/{kind}/{id}/original.ext
+    返回 (relative_path, absolute_dest_path)
+    """
+    sub = media_subdir(kind, media_id)
+    dest_dir = os.path.join(skill_data_dir, sub.replace("/", os.sep))
+    os.makedirs(dest_dir, exist_ok=True)
+    base = original_basename(src_path)
+    abs_dest = os.path.join(dest_dir, base)
+    shutil.copy2(src_path, abs_dest)
+    rel = f"{kind}/{media_id}/{base}".replace("\\", "/")
+    return rel, abs_dest
+
+
+def remove_files_for_relative_path(skill_data_dir: str, relative_file_path: str) -> None:
+    """删除 relative_file_path 所在目录（整 id 目录）。"""
+    rel = (relative_file_path or "").strip().replace("\\", "/")
+    if not rel or "/" not in rel:
+        return
+    parts = rel.split("/")
+    if len(parts) < 2:
+        return
+    id_dir = os.path.join(skill_data_dir, parts[0], parts[1])
+    if os.path.isdir(id_dir):
+        shutil.rmtree(id_dir, ignore_errors=True)

content-manager/content_manager/services/image_service.py

@@ -0,0 +1,171 @@
+"""图片：业务规则（文件落盘 + 路径写入 images 表）。"""
+
+from __future__ import annotations
+
+import json
+import os
+import sys
+from typing import Any, Dict, Optional
+
+from content_manager.config import get_skill_data_dir, resolve_stored_path
+from content_manager.db import images_repository as ir
+from content_manager.db.connection import get_conn, init_db
+from content_manager.services import file_store
+from content_manager.util.timeutil import now_unix, unix_to_iso
+
+
+def _row_to_public_dict(row: tuple) -> Dict[str, Any]:
+    rid, file_path, title, status, source, account_id, error_msg, extra_json, cat, uat = row
+    abs_path = resolve_stored_path(str(file_path))
+    d: Dict[str, Any] = {
+        "id": int(rid),
+        "kind": "image",
+        "file_path": file_path,
+        "absolute_path": abs_path,
+        "title": title,
+        "status": status or "draft",
+        "source": source or "manual",
+        "account_id": account_id,
+        "error_msg": error_msg,
+        "created_at": unix_to_iso(cat),
+        "updated_at": unix_to_iso(uat),
+    }
+    if extra_json:
+        try:
+            ex = json.loads(extra_json)
+            if isinstance(ex, dict):
+                d["extra"] = ex
+        except json.JSONDecodeError:
+            pass
+    return d
+
+
+def cmd_add(src_file: str, title: Optional[str] = None, source: str = "manual") -> None:
+    init_db()
+    src_file = os.path.abspath(src_file.strip())
+    if not os.path.isfile(src_file):
+        print(f"❌ 找不到文件：{src_file}")
+        sys.exit(1)
+    skill_data = get_skill_data_dir()
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        new_id = ir.insert_row(
+            conn,
+            file_path="",
+            title=(title or "").strip() or None,
+            status="draft",
+            source=source,
+            account_id=None,
+            error_msg=None,
+            extra_json=None,
+            created_at=ts,
+            updated_at=ts,
+        )
+        rel, _abs = file_store.copy_into_skill_data(skill_data, "images", new_id, src_file)
+        ir.update_file_path(conn, new_id, rel, now_unix())
+        conn.commit()
+    except Exception:
+        conn.rollback()
+        raise
+    finally:
+        conn.close()
+    print(f"✅ 已新增图片 id={new_id} | 路径：{rel}")
+
+
+def cmd_get(image_id: str) -> None:
+    init_db()
+    if not str(image_id).strip().isdigit():
+        print("❌ 图片 id 必须是纯数字。请先 image list 查看。")
+        sys.exit(1)
+    iid = int(image_id)
+    conn = get_conn()
+    try:
+        row = ir.fetch_by_id(conn, iid)
+    finally:
+        conn.close()
+    if not row:
+        print("❌ 没有这条图片记录。")
+        sys.exit(1)
+    print(json.dumps(_row_to_public_dict(row), ensure_ascii=False))
+
+
+def cmd_list(limit: int = 20, max_chars: int = 80) -> None:
+    init_db()
+    conn = get_conn()
+    try:
+        rows = ir.list_recent(conn, limit)
+    finally:
+        conn.close()
+    if not rows:
+        print("暂无图片")
+        return
+
+    def trunc(s: str) -> str:
+        if not s:
+            return ""
+        return s if len(s) <= max_chars else s[:max_chars] + "..."
+
+    sep = "_" * 39
+    for idx, r in enumerate(rows):
+        rid, file_path, title, status, source, account_id, error_msg, extra_json, cat, uat = r
+        print(f"id：{rid}")
+        print(f"file_path：{trunc(str(file_path or ''))}")
+        print(f"title：{title or ''}")
+        print(f"status：{status or ''}")
+        print(f"source：{source or ''}")
+        print(f"account_id：{account_id or ''}")
+        print(f"error_msg：{error_msg or ''}")
+        print(f"created_at：{unix_to_iso(cat) or ''}")
+        print(f"updated_at：{unix_to_iso(uat) or ''}")
+        if idx != len(rows) - 1:
+            print(sep)
+            print()
+
+
+def cmd_delete(image_id: str) -> None:
+    init_db()
+    if not str(image_id).strip().isdigit():
+        print("❌ 图片 id 必须是纯数字。")
+        sys.exit(1)
+    iid = int(image_id)
+    skill_data = get_skill_data_dir()
+    conn = get_conn()
+    try:
+        row = ir.fetch_by_id(conn, iid)
+        if not row:
+            print("❌ 没有 id 为 {} 的图片记录。".format(iid))
+            sys.exit(1)
+        rel = row[1]
+        n = ir.delete_by_id(conn, iid)
+        if n == 0:
+            sys.exit(1)
+        conn.commit()
+    finally:
+        conn.close()
+    file_store.remove_files_for_relative_path(skill_data, str(rel))
+    print(f"✅ 已删除图片 id={iid}")
+
+
+def cmd_feedback(
+    image_id: str,
+    status: str,
+    account_id: Optional[str] = None,
+    error_msg: Optional[str] = None,
+) -> None:
+    init_db()
+    if not str(image_id).strip().isdigit():
+        print("❌ 图片 id 必须是纯数字。")
+        sys.exit(1)
+    iid = int(image_id)
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        if ir.fetch_by_id(conn, iid) is None:
+            print("❌ 没有 id 为 {} 的图片记录。".format(iid))
+            sys.exit(1)
+        ir.update_feedback(conn, iid, status, account_id, error_msg, ts)
+        conn.commit()
+    finally:
+        conn.close()
+    print("✅ 状态已更新")

content-manager/content_manager/services/video_service.py

@@ -0,0 +1,179 @@
+"""视频：业务规则（文件落盘 + 路径写入 videos 表）。"""
+
+from __future__ import annotations
+
+import json
+import os
+import sys
+from typing import Any, Dict, Optional
+
+from content_manager.config import get_skill_data_dir, resolve_stored_path
+from content_manager.db import videos_repository as vr
+from content_manager.db.connection import get_conn, init_db
+from content_manager.services import file_store
+from content_manager.util.timeutil import now_unix, unix_to_iso
+
+
+def _row_to_public_dict(row: tuple) -> Dict[str, Any]:
+    rid, file_path, title, duration_ms, status, source, account_id, error_msg, extra_json, cat, uat = row
+    abs_path = resolve_stored_path(str(file_path))
+    d: Dict[str, Any] = {
+        "id": int(rid),
+        "kind": "video",
+        "file_path": file_path,
+        "absolute_path": abs_path,
+        "title": title,
+        "duration_ms": duration_ms,
+        "status": status or "draft",
+        "source": source or "manual",
+        "account_id": account_id,
+        "error_msg": error_msg,
+        "created_at": unix_to_iso(cat),
+        "updated_at": unix_to_iso(uat),
+    }
+    if extra_json:
+        try:
+            ex = json.loads(extra_json)
+            if isinstance(ex, dict):
+                d["extra"] = ex
+        except json.JSONDecodeError:
+            pass
+    return d
+
+
+def cmd_add(
+    src_file: str,
+    title: Optional[str] = None,
+    duration_ms: Optional[int] = None,
+    source: str = "manual",
+) -> None:
+    init_db()
+    src_file = os.path.abspath(src_file.strip())
+    if not os.path.isfile(src_file):
+        print(f"❌ 找不到文件：{src_file}")
+        sys.exit(1)
+    skill_data = get_skill_data_dir()
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        new_id = vr.insert_row(
+            conn,
+            file_path="",
+            title=(title or "").strip() or None,
+            duration_ms=duration_ms,
+            status="draft",
+            source=source,
+            account_id=None,
+            error_msg=None,
+            extra_json=None,
+            created_at=ts,
+            updated_at=ts,
+        )
+        rel, _abs = file_store.copy_into_skill_data(skill_data, "videos", new_id, src_file)
+        vr.update_file_path(conn, new_id, rel, now_unix())
+        conn.commit()
+    except Exception:
+        conn.rollback()
+        raise
+    finally:
+        conn.close()
+    print(f"✅ 已新增视频 id={new_id} | 路径：{rel}")
+
+
+def cmd_get(video_id: str) -> None:
+    init_db()
+    if not str(video_id).strip().isdigit():
+        print("❌ 视频 id 必须是纯数字。请先 video list 查看。")
+        sys.exit(1)
+    vid = int(video_id)
+    conn = get_conn()
+    try:
+        row = vr.fetch_by_id(conn, vid)
+    finally:
+        conn.close()
+    if not row:
+        print("❌ 没有这条视频记录。")
+        sys.exit(1)
+    print(json.dumps(_row_to_public_dict(row), ensure_ascii=False))
+
+
+def cmd_list(limit: int = 20, max_chars: int = 80) -> None:
+    init_db()
+    conn = get_conn()
+    try:
+        rows = vr.list_recent(conn, limit)
+    finally:
+        conn.close()
+    if not rows:
+        print("暂无视频")
+        return
+
+    def trunc(s: str) -> str:
+        if not s:
+            return ""
+        return s if len(s) <= max_chars else s[:max_chars] + "..."
+
+    sep = "_" * 39
+    for idx, r in enumerate(rows):
+        rid, file_path, title, duration_ms, status, source, account_id, error_msg, extra_json, cat, uat = r
+        print(f"id：{rid}")
+        print(f"file_path：{trunc(str(file_path or ''))}")
+        print(f"title：{title or ''}")
+        print(f"duration_ms：{duration_ms if duration_ms is not None else ''}")
+        print(f"status：{status or ''}")
+        print(f"source：{source or ''}")
+        print(f"account_id：{account_id or ''}")
+        print(f"error_msg：{error_msg or ''}")
+        print(f"created_at：{unix_to_iso(cat) or ''}")
+        print(f"updated_at：{unix_to_iso(uat) or ''}")
+        if idx != len(rows) - 1:
+            print(sep)
+            print()
+
+
+def cmd_delete(video_id: str) -> None:
+    init_db()
+    if not str(video_id).strip().isdigit():
+        print("❌ 视频 id 必须是纯数字。")
+        sys.exit(1)
+    vid = int(video_id)
+    skill_data = get_skill_data_dir()
+    conn = get_conn()
+    try:
+        row = vr.fetch_by_id(conn, vid)
+        if not row:
+            print("❌ 没有 id 为 {} 的视频记录。".format(vid))
+            sys.exit(1)
+        rel = row[1]
+        n = vr.delete_by_id(conn, vid)
+        if n == 0:
+            sys.exit(1)
+        conn.commit()
+    finally:
+        conn.close()
+    file_store.remove_files_for_relative_path(skill_data, str(rel))
+    print(f"✅ 已删除视频 id={vid}")
+
+
+def cmd_feedback(
+    video_id: str,
+    status: str,
+    account_id: Optional[str] = None,
+    error_msg: Optional[str] = None,
+) -> None:
+    init_db()
+    if not str(video_id).strip().isdigit():
+        print("❌ 视频 id 必须是纯数字。")
+        sys.exit(1)
+    vid = int(video_id)
+    ts = now_unix()
+    conn = get_conn()
+    try:
+        if vr.fetch_by_id(conn, vid) is None:
+            print("❌ 没有 id 为 {} 的视频记录。".format(vid))
+            sys.exit(1)
+        vr.update_feedback(conn, vid, status, account_id, error_msg, ts)
+        conn.commit()
+    finally:
+        conn.close()
+    print("✅ 状态已更新")

content-manager/content_manager/util/__init__.py

@@ -0,0 +1 @@
+# 工具函数

content-manager/content_manager/util/argparse_zh.py

@@ -0,0 +1,74 @@
+"""argparse 中文错误说明。"""
+
+from __future__ import annotations
+
+import argparse
+import sys
+from typing import List
+
+from content_manager.constants import CLI_REQUIRED_ZH
+
+
+def split_required_arg_names(raw: str) -> List[str]:
+    s = raw.replace(" and ", ", ").strip()
+    parts: List[str] = []
+    for chunk in s.split(","):
+        chunk = chunk.strip()
+        if not chunk:
+            continue
+        idx = chunk.find(" --")
+        if idx != -1:
+            left = chunk[:idx].strip()
+            flag_rest = chunk[idx + 1 :].strip().split()
+            if left:
+                parts.append(left)
+            if flag_rest:
+                parts.append(flag_rest[0])
+        else:
+            parts.append(chunk)
+    return [p for p in parts if p]
+
+
+def explain_argparse_error(message: str) -> str:
+    m = (message or "").strip()
+    lines: List[str] = ["【命令参数不完整或写错了】请对照下面修改后再执行。"]
+    if "the following arguments are required:" in m:
+        raw = m.split("required:", 1)[-1].strip()
+        parts = split_required_arg_names(raw)
+        for p in parts:
+            hint = CLI_REQUIRED_ZH.get(p)
+            if not hint and p.startswith("--"):
+                hint = CLI_REQUIRED_ZH.get(p.split()[0], None)
+            lines.append(f"  · {hint or f'还缺这一项：{p}'}")
+        lines.append("  · 查看全部：python main.py -h")
+        lines.append("  · 查看分组：python main.py article -h")
+        return "\n".join(lines)
+    if "one of the arguments" in m and "required" in m:
+        lines.append("  · 本命令要求下面几组参数里「必须选其中一组」，不能都不写。")
+        if "--body-file" in m and "--body" in m:
+            lines.append("  · 请任选其一：--body-file 某文件路径   或   --body \"正文文字\"")
+        else:
+            lines.append(f"  · 说明：{m}")
+        lines.append("  · 查看该子命令：python main.py article add -h")
+        return "\n".join(lines)
+    if "unrecognized arguments:" in m:
+        tail = m.split("unrecognized arguments:", 1)[-1].strip()
+        lines.append(f"  · 多写了不认识的参数：{tail}，请删除或检查拼写。")
+        lines.append("  · 查看用法：python main.py -h")
+        return "\n".join(lines)
+    if "invalid choice:" in m:
+        lines.append(f"  · {m}")
+        return "\n".join(lines)
+    if "expected one argument" in m:
+        lines.append(f"  · {m}")
+        lines.append("  · 提示：--xxx 后面必须跟一个值，不要忘记。")
+        return "\n".join(lines)
+    lines.append(f"  · {m}")
+    lines.append("  · 查看帮助：python main.py -h")
+    return "\n".join(lines)
+
+
+class ZhArgumentParser(argparse.ArgumentParser):
+    def error(self, message: str) -> None:
+        print(explain_argparse_error(message), file=sys.stderr)
+        self.exit(2)

content-manager/content_manager/util/timeutil.py

@@ -0,0 +1,36 @@
+"""时间戳工具。"""
+
+from __future__ import annotations
+
+import time
+from datetime import datetime
+from typing import Any, Optional
+
+
+def now_unix() -> int:
+    return int(time.time())
+
+
+def unix_to_iso(ts: Optional[int]) -> Optional[str]:
+    if ts is None:
+        return None
+    try:
+        return datetime.fromtimestamp(int(ts)).isoformat(timespec="seconds")
+    except (ValueError, OSError, OverflowError):
+        return None
+
+
+def parse_ts_to_unix(val: Any) -> Optional[int]:
+    if val is None:
+        return None
+    if isinstance(val, (int, float)):
+        return int(val)
+    s = str(val).strip()
+    if not s:
+        return None
+    if s.isdigit():
+        return int(s)
+    try:
+        return int(datetime.fromisoformat(s.replace("Z", "+00:00")).timestamp())
+    except ValueError:
+        return None

content-manager/release.ps1

@@ -0,0 +1,220 @@
+<#
+.SYNOPSIS
+  One-command release script for skill repos.
+
+.DESCRIPTION
+  - Optional auto-commit
+  - Push current branch
+  - Auto-increment semantic tag (vX.Y.Z)
+  - Create & push tag
+  - Fail fast on unsafe states
+
+.EXAMPLES
+  # Safe mode (recommended): requires clean working tree
+  .\release.ps1
+
+  # Auto commit tracked/untracked changes then release
+  .\release.ps1 -AutoCommit -CommitMessage "chore: update skill config"
+
+  # Dry run (show what would happen)
+  .\release.ps1 -DryRun
+
+  # Custom tag prefix
+  .\release.ps1 -Prefix "v" -Message "正式发布"
+
+.NOTES
+  Requires: git, PowerShell 5+
+#>
+
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+if (Test-Path $sharedScript) {
+  & $sharedScript @PSBoundParameters
+  exit $LASTEXITCODE
+}
+
+function Invoke-Git {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  Write-Host ">> git $Args" -ForegroundColor DarkGray
+  & cmd /c "git $Args"
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+}
+
+function Get-GitOutput {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  $output = & cmd /c "git $Args" 2>$null
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+  return @($output)
+}
+
+function Test-Repo {
+  & git rev-parse --is-inside-work-tree *> $null
+  return ($LASTEXITCODE -eq 0)
+}
+
+function Get-CurrentBranch {
+  $b = (Get-GitOutput "branch --show-current" | Select-Object -First 1).Trim()
+  return $b
+}
+
+function Get-StatusPorcelain {
+  $lines = @(Get-GitOutput "status --porcelain")
+  return $lines
+}
+
+function Parse-SemVerTag {
+  param(
+    [string]$Tag,
+    [string]$TagPrefix
+  )
+  $escaped = [regex]::Escape($TagPrefix)
+  $m = [regex]::Match($Tag, "^${escaped}(\d+)\.(\d+)\.(\d+)$")
+  if (-not $m.Success) { return $null }
+
+  return [pscustomobject]@{
+    Raw   = $Tag
+    Major = [int]$m.Groups[1].Value
+    Minor = [int]$m.Groups[2].Value
+    Patch = [int]$m.Groups[3].Value
+  }
+}
+
+function Get-NextTag {
+  param([string]$TagPrefix)
+
+  $tags = Get-GitOutput "tag --list"
+  $parsed = @()
+
+  foreach ($t in $tags) {
+    $t = $t.Trim()
+    if (-not $t) { continue }
+    $obj = Parse-SemVerTag -Tag $t -TagPrefix $TagPrefix
+    if ($null -ne $obj) { $parsed += $obj }
+  }
+
+  if ($parsed.Count -eq 0) {
+    return "${TagPrefix}1.0.1"
+  }
+
+  $latest = $parsed | Sort-Object Major, Minor, Patch | Select-Object -Last 1
+  return "$TagPrefix$($latest.Major).$($latest.Minor).$([int]$latest.Patch + 1)"
+}
+
+function Ensure-CleanOrAutoCommit {
+  param(
+    [switch]$DoAutoCommit,
+    [switch]$NeedClean,
+    [switch]$IsDryRun,
+    [string]$Msg
+  )
+
+  $status = @(Get-StatusPorcelain)
+  if ($status.Length -eq 0) { return }
+
+  if ($NeedClean) {
+    Write-Host "Working tree is not clean and -RequireClean is enabled." -ForegroundColor Yellow
+    & git status --short
+    throw "Abort: dirty working tree."
+  }
+
+  # 默认一键发布：有改动就自动提交；也可用 -AutoCommit 显式开启
+  $commitMsg = $Msg
+  if ([string]::IsNullOrWhiteSpace($commitMsg)) {
+    $commitMsg = "chore: auto release commit ($(Get-Date -Format 'yyyy-MM-dd HH:mm:ss'))"
+  }
+
+  if (-not $DoAutoCommit) {
+    Write-Host "Detected uncommitted changes, auto-committing before release..." -ForegroundColor Yellow
+  }
+
+  if ($IsDryRun) {
+    Write-Host "[DryRun] Would run: git add -A" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git commit -m `"$commitMsg`"" -ForegroundColor Yellow
+    return
+  }
+
+  Invoke-Git "add -A"
+  Invoke-Git "commit -m `"$commitMsg`""
+}
+
+try {
+  Write-Host "=== Release Script Start ===" -ForegroundColor Cyan
+
+  if (-not (Test-Repo)) {
+    throw "Current directory is not a git repository."
+  }
+
+  $branch = Get-CurrentBranch
+  if ([string]::IsNullOrWhiteSpace($branch)) {
+    throw "Unable to determine current branch."
+  }
+
+  if ($branch -notin @("main", "master")) {
+    throw "Current branch is '$branch'. Release is only allowed from main/master."
+  }
+
+  Invoke-Git "fetch --tags --prune origin"
+
+  Ensure-CleanOrAutoCommit -DoAutoCommit:$AutoCommit -NeedClean:$RequireClean -IsDryRun:$DryRun -Msg $CommitMessage
+
+  $upstream = (& git rev-parse --abbrev-ref --symbolic-full-name "@{u}" 2>$null)
+  $hasUpstream = ($LASTEXITCODE -eq 0)
+
+  if ($DryRun) {
+    if ($hasUpstream) {
+      Write-Host "[DryRun] Would run: git push" -ForegroundColor Yellow
+    } else {
+      Write-Host "[DryRun] Would run: git push -u origin $branch" -ForegroundColor Yellow
+    }
+  } else {
+    if ($hasUpstream) {
+      Invoke-Git "push"
+    } else {
+      Invoke-Git "push -u origin $branch"
+    }
+  }
+
+  $nextTag = Get-NextTag -TagPrefix $Prefix
+  Write-Host "Next tag: $nextTag" -ForegroundColor Green
+
+  $existing = @(Get-GitOutput "tag --list `"$nextTag`"")
+  if ($existing.Length -gt 0) {
+    throw "Tag already exists: $nextTag"
+  }
+
+  if ($DryRun) {
+    Write-Host "[DryRun] Would run: git tag -a $nextTag -m `"$Message`"" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git push origin $nextTag" -ForegroundColor Yellow
+    Write-Host "=== DryRun Complete ===" -ForegroundColor Cyan
+    exit 0
+  }
+
+  Invoke-Git "tag -a $nextTag -m `"$Message`""
+  Invoke-Git "push origin $nextTag"
+
+  Write-Host "Release success: $nextTag" -ForegroundColor Green
+  Write-Host "=== Release Script Done ===" -ForegroundColor Cyan
+  exit 0
+}
+catch {
+  Write-Host "Release failed: $($_.Exception.Message)" -ForegroundColor Red
+  exit 1
+}

content-manager/scripts/main.py

@@ -0,0 +1,24 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""content-manager CLI 入口；逻辑在 content_manager 包内分层实现。"""
+
+from __future__ import annotations
+
+import os
+import sys
+
+# Windows GBK 下控制台 UTF-8 输出
+if sys.platform == "win32":
+    import io
+
+    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
+    sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace")
+
+_SKILL_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if _SKILL_ROOT not in sys.path:
+    sys.path.insert(0, _SKILL_ROOT)
+
+from content_manager.cli.app import main
+
+if __name__ == "__main__":
+    raise SystemExit(main())

jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml

@@ -0,0 +1,139 @@
+name: Reusable Skill Release
+
+on:
+  workflow_call:
+    inputs:
+      artifact_platform:
+        required: false
+        type: string
+        default: windows
+      pyarmor_platform:
+        required: false
+        type: string
+        default: windows.x86_64
+      upload_url:
+        required: false
+        type: string
+        default: https://jc2009.com/api/upload
+      sync_url:
+        required: false
+        type: string
+        default: https://jc2009.com/api/skill/update
+      prune_url:
+        required: false
+        type: string
+        default: https://jc2009.com/api/artifacts/prune-old-versions
+
+jobs:
+  build-and-deploy:
+    runs-on: ubuntu-latest
+    env:
+      ARTIFACT_PLATFORM: ${{ inputs.artifact_platform }}
+      PYARMOR_PLATFORM: ${{ inputs.pyarmor_platform }}
+      PIP_BREAK_SYSTEM_PACKAGES: "1"
+    steps:
+      - uses: http://120.25.191.12:3000/admin/actions-checkout@v4
+
+      - name: Setup Tools
+        run: pip install "pyarmor>=8.5" requests python-frontmatter --break-system-packages -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+      - name: Encrypt Source Code
+        run: |
+          mkdir -p dist/package
+          pyarmor gen --platform "${PYARMOR_PLATFORM}" -O dist/package scripts/*.py
+          cp SKILL.md dist/package/
+
+      - name: Parse Metadata and Pack
+        id: build_task
+        run: |
+          python -c "
+          import frontmatter, os, json, shutil
+          post = frontmatter.load('SKILL.md')
+          metadata = dict(post.metadata or {})
+          metadata['readme_md'] = (post.content or '').strip()
+          openclaw_meta = metadata.get('metadata', {}).get('openclaw', {})
+          slug = (openclaw_meta.get('slug') or metadata.get('slug') or metadata.get('name') or '').strip()
+          if not slug:
+              raise Exception('SKILL.md 缺少 slug/name')
+          ref_name = (os.environ.get('GITHUB_REF_NAME') or '').strip()
+          if not ref_name.startswith('v'):
+              raise Exception(f'非法标签: {ref_name}')
+          version = ref_name.lstrip('v')
+          metadata['version'] = version
+          artifact_platform = (os.environ.get('ARTIFACT_PLATFORM') or 'windows').strip()
+          zip_base = f'{slug}-{artifact_platform}'
+          with open(os.environ['GITHUB_OUTPUT'], 'a') as f:
+              f.write(f'slug={slug}\n')
+              f.write(f'version={version}\n')
+              f.write(f'zip_base={zip_base}\n')
+              f.write(f'artifact_platform={artifact_platform}\n')
+              f.write(f'metadata={json.dumps(metadata, ensure_ascii=False)}\n')
+          shutil.make_archive(zip_base, 'zip', 'dist/package')
+          "
+
+      - name: Sync Database
+        env:
+          METADATA_JSON: ${{ steps.build_task.outputs.metadata }}
+          SYNC_URL: ${{ inputs.sync_url }}
+        run: |
+          python -c "
+          import requests, json, os
+          metadata = json.loads(os.environ['METADATA_JSON'])
+          res = requests.post(os.environ['SYNC_URL'], json=metadata)
+          if res.status_code != 200:
+              exit(1)
+          body = res.json()
+          if body.get('code') != 200:
+              exit(1)
+          "
+
+      - name: Upload Encrypted ZIP
+        env:
+          SLUG: ${{ steps.build_task.outputs.slug }}
+          VERSION: ${{ steps.build_task.outputs.version }}
+          ZIP_BASE: ${{ steps.build_task.outputs.zip_base }}
+          ARTIFACT_PLATFORM: ${{ steps.build_task.outputs.artifact_platform }}
+          UPLOAD_URL: ${{ inputs.upload_url }}
+        run: |
+          python -c "
+          import requests, os
+          slug = os.environ['SLUG']
+          version = os.environ['VERSION']
+          zip_path = f\"{os.environ['ZIP_BASE']}.zip\"
+          payload = {
+              'plugin_name': slug,
+              'version': version,
+              'artifact_type': 'skill',
+              'artifact_platform': os.environ.get('ARTIFACT_PLATFORM', 'windows'),
+          }
+          filename = os.path.basename(zip_path)
+          with open(zip_path, 'rb') as f:
+              res = requests.post(os.environ['UPLOAD_URL'], data=payload, files={'file': (filename, f)})
+          if res.status_code != 200:
+              exit(1)
+          body = res.json()
+          if body.get('code') != 200:
+              exit(1)
+          "
+
+      - name: Prune Old Versions
+        env:
+          SLUG: ${{ steps.build_task.outputs.slug }}
+          VERSION: ${{ steps.build_task.outputs.version }}
+          PRUNE_URL: ${{ inputs.prune_url }}
+        run: |
+          python -c "
+          import requests, os
+          payload = {
+            'name': os.environ['SLUG'],
+            'artifact_type': 'skill',
+            'keep_count': 1,
+            'protect_version': os.environ['VERSION']
+          }
+          res = requests.post(os.environ['PRUNE_URL'], json=payload)
+          if res.status_code != 200:
+              exit(1)
+          body = res.json()
+          if body.get('code') != 200:
+              exit(1)
+          "

jiangchang-platform-kit/README.md

@@ -0,0 +1,7 @@
+# jiangchang-platform-kit
+
+Shared platform components for Jiangchang skills:
+
+- `sdk/jiangchang_skill_core`: entitlement SDK package.
+- `.github/workflows/reusable-release-skill.yaml`: reusable CI release workflow.
+- `examples/workflows/use-reusable-release-skill.yaml`: caller workflow sample.

jiangchang-platform-kit/examples/flask_skill_update_or_create.py

@@ -0,0 +1,84 @@
+# 将此路由合并进你的 skill 蓝图。
+# CI 不再写入 skill_type / monthly_price / yearly_price，避免每次发布覆盖后台手工配置。
+#
+# 要求：SkillModel.update_or_create 在「更新」时对 data 中未出现的列应保留数据库原值；
+# 若当前是整行覆盖，请在 Model 层改为按字段合并或白名单更新。
+# 若表对这三列 NOT NULL 且无默认值，仅在「首次插入」时在 Model 内写死默认即可。
+
+import json
+import re
+from datetime import datetime
+
+from flask import jsonify, request
+
+# from your_app import skill_bp, SkillModel  # 按你项目实际导入
+
+
+@skill_bp.route("/api/skill/update", methods=["POST"])
+def update_or_create_skill():
+    """CI/CD 自动化注册接口（可选携带 readme_md = SKILL.md 正文 Markdown）"""
+    try:
+        data = request.get_json(silent=True) or {}
+        if not data:
+            return jsonify({"code": 400, "msg": "请求体为空", "data": None}), 400
+
+        openclaw_meta = data.get("metadata", {}).get("openclaw", {})
+
+        slug = (
+            (data.get("slug") or "").strip()
+            or (openclaw_meta.get("slug") or "").strip()
+            or (data.get("name") or "").strip()
+        )
+        name = (data.get("name") or slug).strip()
+        version = str(data.get("version") or "").strip()
+        category = (openclaw_meta.get("category") or "").strip()
+
+        if not all([slug, name, version, category]):
+            return jsonify(
+                {
+                    "code": 400,
+                    "msg": "元数据不完整（需包含 slug/name/version/category）",
+                    "data": None,
+                }
+            ), 400
+
+        slug = slug.lower()
+
+        if not re.match(r"^[a-z0-9-]+$", slug):
+            return jsonify(
+                {
+                    "code": 400,
+                    "msg": "slug 格式非法，仅允许小写字母、数字和中划线",
+                    "data": None,
+                }
+            ), 400
+
+        skill_data = {
+            "slug": slug,
+            "name": name,
+            "description": data.get("description"),
+            "version": version,
+            "category": category,
+            "developer_name": data.get("author", "匠厂开发者"),
+            "tags": json.dumps(data.get("tags", []), ensure_ascii=False),
+            "status": 2,
+            "updated_at": datetime.now(),
+        }
+
+        if "readme_md" in data:
+            rm = data.get("readme_md")
+            skill_data["readme_md"] = "" if rm is None else str(rm)
+
+        success = SkillModel.update_or_create(slug=slug, data=skill_data)
+        if success:
+            return jsonify(
+                {
+                    "code": 200,
+                    "msg": "注册成功",
+                    "data": {"slug": slug, "name": name, "version": version},
+                }
+            ), 200
+        return jsonify({"code": 500, "msg": "数据持久化失败", "data": None}), 500
+
+    except Exception as e:
+        return jsonify({"code": 500, "msg": str(e), "data": None}), 500

jiangchang-platform-kit/examples/workflows/use-reusable-release-skill.yaml

@@ -0,0 +1,12 @@
+name: Skill Release
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@v0.1.0
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64
+      include_readme_md: false

jiangchang-platform-kit/sdk/jiangchang_skill_core/__init__.py

@@ -0,0 +1,9 @@
+from .client import EntitlementClient
+from .guard import enforce_entitlement
+from .models import EntitlementResult
+
+__all__ = [
+    "EntitlementClient",
+    "EntitlementResult",
+    "enforce_entitlement",
+]

jiangchang-platform-kit/sdk/jiangchang_skill_core/client.py

@@ -0,0 +1,63 @@
+import os
+from typing import Any
+
+import requests
+
+from .errors import EntitlementServiceError
+from .models import EntitlementResult
+
+
+class EntitlementClient:
+    def __init__(
+        self,
+        base_url: str | None = None,
+        api_key: str | None = None,
+        timeout_seconds: int | None = None,
+    ) -> None:
+        self.base_url = (base_url or os.getenv("JIANGCHANG_AUTH_BASE_URL", "")).rstrip("/")
+        self.api_key = api_key or os.getenv("JIANGCHANG_AUTH_API_KEY", "")
+        self.timeout_seconds = timeout_seconds or int(
+            os.getenv("JIANGCHANG_AUTH_TIMEOUT_SECONDS", "5")
+        )
+        if not self.base_url:
+            raise EntitlementServiceError("missing JIANGCHANG_AUTH_BASE_URL")
+
+    def check_entitlement(
+        self,
+        user_id: str,
+        skill_slug: str,
+        trace_id: str = "",
+        context: dict[str, Any] | None = None,
+    ) -> EntitlementResult:
+        url = f"{self.base_url}/api/entitlements/check"
+        payload = {
+            "user_id": user_id,
+            "skill_slug": skill_slug,
+            "trace_id": trace_id,
+            "context": context or {},
+        }
+        headers = {"Content-Type": "application/json"}
+        if self.api_key:
+            headers["Authorization"] = f"Bearer {self.api_key}"
+        try:
+            res = requests.post(
+                url,
+                json=payload,
+                headers=headers,
+                timeout=self.timeout_seconds,
+            )
+        except requests.RequestException as exc:
+            raise EntitlementServiceError(f"entitlement request failed: {exc}") from exc
+
+        if res.status_code != 200:
+            raise EntitlementServiceError(f"entitlement http status {res.status_code}")
+        try:
+            body = res.json()
+        except ValueError as exc:
+            raise EntitlementServiceError("entitlement response is not json") from exc
+
+        data = body.get("data") or {}
+        allow = bool(data.get("allow", False))
+        reason = str(data.get("reason") or body.get("msg") or "")
+        expire_at = str(data.get("expire_at") or "")
+        return EntitlementResult(allow=allow, reason=reason, expire_at=expire_at, raw=body)

jiangchang-platform-kit/sdk/jiangchang_skill_core/errors.py

@@ -0,0 +1,10 @@
+class EntitlementError(Exception):
+    pass
+
+
+class EntitlementDeniedError(EntitlementError):
+    pass
+
+
+class EntitlementServiceError(EntitlementError):
+    pass

jiangchang-platform-kit/sdk/jiangchang_skill_core/guard.py

@@ -0,0 +1,22 @@
+from .client import EntitlementClient
+from .errors import EntitlementDeniedError
+from .models import EntitlementResult
+
+
+def enforce_entitlement(
+    user_id: str,
+    skill_slug: str,
+    trace_id: str = "",
+    context: dict | None = None,
+    client: EntitlementClient | None = None,
+) -> EntitlementResult:
+    c = client or EntitlementClient()
+    result = c.check_entitlement(
+        user_id=user_id,
+        skill_slug=skill_slug,
+        trace_id=trace_id,
+        context=context or {},
+    )
+    if not result.allow:
+        raise EntitlementDeniedError(result.reason or "skill not purchased or expired")
+    return result

jiangchang-platform-kit/sdk/jiangchang_skill_core/models.py

@@ -0,0 +1,10 @@
+from dataclasses import dataclass
+from typing import Any
+
+
+@dataclass
+class EntitlementResult:
+    allow: bool
+    reason: str = ""
+    expire_at: str = ""
+    raw: dict[str, Any] | None = None

jiangchang-platform-kit/sdk/pyproject.toml

@@ -0,0 +1,19 @@
+[build-system]
+requires = ["setuptools>=68", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "jiangchang-skill-core"
+version = "0.1.0"
+description = "Common entitlement SDK for Jiangchang skills"
+requires-python = ">=3.10"
+dependencies = [
+  "requests>=2.31.0",
+]
+
+[tool.setuptools]
+package-dir = {"" = "."}
+
+[tool.setuptools.packages.find]
+where = ["."]
+include = ["jiangchang_skill_core*"]

jiangchang-platform-kit/tools/release.ps1

@@ -0,0 +1,216 @@
+<#
+.SYNOPSIS
+  One-command release script for skill repos.
+
+.DESCRIPTION
+  - Optional auto-commit
+  - Push current branch
+  - Auto-increment semantic tag (vX.Y.Z)
+  - Create & push tag
+  - Fail fast on unsafe states
+
+.EXAMPLES
+  # Safe mode (recommended): requires clean working tree
+  .\release.ps1
+
+  # Auto commit tracked/untracked changes then release
+  .\release.ps1 -AutoCommit -CommitMessage "chore: update skill config"
+
+  # Dry run (show what would happen)
+  .\release.ps1 -DryRun
+
+  # Custom tag prefix
+  .\release.ps1 -Prefix "v" -Message "正式发布"
+
+.NOTES
+  Requires: git, PowerShell 5+
+#>
+
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+function Invoke-Git {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  Write-Host ">> git $Args" -ForegroundColor DarkGray
+  & cmd /c "git $Args"
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+}
+
+
+function Get-GitOutput {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  $output = & cmd /c "git $Args" 2>$null
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+  return @($output)
+}
+
+function Test-Repo {
+  & git rev-parse --is-inside-work-tree *> $null
+  return ($LASTEXITCODE -eq 0)
+}
+
+function Get-CurrentBranch {
+  $b = (Get-GitOutput "branch --show-current" | Select-Object -First 1).Trim()
+  return $b
+}
+
+
+function Get-StatusPorcelain {
+  $lines = @(Get-GitOutput "status --porcelain")
+  return $lines
+}
+
+function Parse-SemVerTag {
+  param(
+    [string]$Tag,
+    [string]$TagPrefix
+  )
+  $escaped = [regex]::Escape($TagPrefix)
+  $m = [regex]::Match($Tag, "^${escaped}(\d+)\.(\d+)\.(\d+)$")
+  if (-not $m.Success) { return $null }
+
+  return [pscustomobject]@{
+    Raw   = $Tag
+    Major = [int]$m.Groups[1].Value
+    Minor = [int]$m.Groups[2].Value
+    Patch = [int]$m.Groups[3].Value
+  }
+}
+
+function Get-NextTag {
+  param([string]$TagPrefix)
+
+  $tags = Get-GitOutput "tag --list"
+  $parsed = @()
+
+  foreach ($t in $tags) {
+    $t = $t.Trim()
+    if (-not $t) { continue }
+    $obj = Parse-SemVerTag -Tag $t -TagPrefix $TagPrefix
+    if ($null -ne $obj) { $parsed += $obj }
+  }
+
+  if ($parsed.Count -eq 0) {
+    return "${TagPrefix}1.0.1"
+  }
+
+  $latest = $parsed | Sort-Object Major, Minor, Patch | Select-Object -Last 1
+  return "$TagPrefix$($latest.Major).$($latest.Minor).$([int]$latest.Patch + 1)"
+}
+
+
+function Ensure-CleanOrAutoCommit {
+  param(
+    [switch]$DoAutoCommit,
+    [switch]$NeedClean,
+    [switch]$IsDryRun,
+    [string]$Msg
+  )
+
+  $status = @(Get-StatusPorcelain)
+  if ($status.Length -eq 0) { return }
+
+  if ($NeedClean) {
+    Write-Host "Working tree is not clean and -RequireClean is enabled." -ForegroundColor Yellow
+    & git status --short
+    throw "Abort: dirty working tree."
+  }
+
+  # 默认一键发布：有改动就自动提交；也可用 -AutoCommit 显式开启
+  $commitMsg = $Msg
+  if ([string]::IsNullOrWhiteSpace($commitMsg)) {
+    $commitMsg = "chore: auto release commit ($(Get-Date -Format 'yyyy-MM-dd HH:mm:ss'))"
+  }
+
+  if (-not $DoAutoCommit) {
+    Write-Host "Detected uncommitted changes, auto-committing before release..." -ForegroundColor Yellow
+  }
+
+  if ($IsDryRun) {
+    Write-Host "[DryRun] Would run: git add -A" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git commit -m `"$commitMsg`"" -ForegroundColor Yellow
+    return
+  }
+
+  Invoke-Git "add -A"
+  Invoke-Git "commit -m `"$commitMsg`""
+}
+
+
+try {
+  Write-Host "=== Release Script Start ===" -ForegroundColor Cyan
+
+  if (-not (Test-Repo)) {
+    throw "Current directory is not a git repository."
+  }
+
+  $branch = Get-CurrentBranch
+  if ([string]::IsNullOrWhiteSpace($branch)) {
+    throw "Unable to determine current branch."
+  }
+
+  if ($branch -notin @("main", "master")) {
+    throw "Current branch is '$branch'. Release is only allowed from main/master."
+  }
+
+  Invoke-Git "fetch --tags --prune origin"
+
+  Ensure-CleanOrAutoCommit -DoAutoCommit:$AutoCommit -NeedClean:$RequireClean -IsDryRun:$DryRun -Msg $CommitMessage
+
+  $upstream = (& git rev-parse --abbrev-ref --symbolic-full-name "@{u}" 2>$null)
+  $hasUpstream = ($LASTEXITCODE -eq 0)
+
+  if ($DryRun) {
+    if ($hasUpstream) {
+      Write-Host "[DryRun] Would run: git push" -ForegroundColor Yellow
+    } else {
+      Write-Host "[DryRun] Would run: git push -u origin $branch" -ForegroundColor Yellow
+    }
+  } else {
+    if ($hasUpstream) {
+      Invoke-Git "push"
+    } else {
+      Invoke-Git "push -u origin $branch"
+    }
+  }
+
+  $nextTag = Get-NextTag -TagPrefix $Prefix
+  Write-Host "Next tag: $nextTag" -ForegroundColor Green
+
+  $existing = @(Get-GitOutput "tag --list `"$nextTag`"")
+  if ($existing.Length -gt 0) {
+    throw "Tag already exists: $nextTag"
+  }
+
+  if ($DryRun) {
+    Write-Host "[DryRun] Would run: git tag -a $nextTag -m `"$Message`"" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git push origin $nextTag" -ForegroundColor Yellow
+    Write-Host "=== DryRun Complete ===" -ForegroundColor Cyan
+    exit 0
+  }
+
+  Invoke-Git "tag -a $nextTag -m `"$Message`""
+  Invoke-Git "push origin $nextTag"
+
+  Write-Host "Release success: $nextTag" -ForegroundColor Green
+  Write-Host "=== Release Script Done ===" -ForegroundColor Cyan
+  exit 0
+}
+catch {
+  Write-Host "Release failed: $($_.Exception.Message)" -ForegroundColor Red
+  exit 1
+}

llm-manager/.github/workflows/release_skill.yaml

@@ -0,0 +1,11 @@
+name: 技能自动化发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

llm-manager/SKILL.md

@@ -0,0 +1,119 @@
+---
+name: 大模型管理器
+description: 统一管理 AI 大模型的调用方式。优先通过账号网页版（免费）访问，无可用网页账号时自动降级到 API Key（付费）。支持豆包、DeepSeek、通义千问、Kimi、文心一言、腾讯元宝。
+version: 1.0.3
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: llm-manager
+    emoji: "🧠"
+    category: "通用"
+    dependencies:
+      required:
+        - account-manager
+      optional:
+        - openai  # pip install openai（仅 API Key 模式需要）
+      auto_install: false
+allowed-tools:
+  - bash
+disable: false
+---
+
+# 大模型管理器
+
+## 使用方式
+
+### 生成内容（核心命令）
+
+```bash
+# 指定平台名（自动选择模式：网页优先，API Key 备用）
+python3 {baseDir}/scripts/main.py generate kimi "帮我写一篇关于AI发展的文章"
+python3 {baseDir}/scripts/main.py generate deepseek "帮我写一篇关于AI发展的文章"
+python3 {baseDir}/scripts/main.py generate doubao "帮我写一篇关于AI发展的文章"
+python3 {baseDir}/scripts/main.py generate qianwen "帮我写一篇关于AI发展的文章"
+python3 {baseDir}/scripts/main.py generate yiyan "帮我写一篇关于AI发展的文章"
+python3 {baseDir}/scripts/main.py generate yuanbao "帮我写一篇关于AI发展的文章"
+
+# 指定 account_id（强制使用该账号网页模式）
+python3 {baseDir}/scripts/main.py generate 3 "帮我写一篇关于AI发展的文章"
+```
+
+### 管理 API Key（可选，付费模式）
+
+```bash
+# 添加 API Key
+python3 {baseDir}/scripts/main.py key add deepseek "sk-xxx"
+python3 {baseDir}/scripts/main.py key add kimi "sk-xxx" --model moonshot-v1-8k
+python3 {baseDir}/scripts/main.py key add doubao "xxx" --model ep-xxx  # 豆包须填 endpoint_id
+
+# 查看
+python3 {baseDir}/scripts/main.py key list
+python3 {baseDir}/scripts/main.py key list deepseek
+
+# 删除
+python3 {baseDir}/scripts/main.py key del <key_id>
+```
+
+### 健康检查
+
+```bash
+python3 {baseDir}/scripts/main.py health
+python3 {baseDir}/scripts/main.py version
+```
+
+## 调用模式选择逻辑
+
+```
+generate <target> "<prompt>"
+  │
+  ├─ target 是数字（account_id）
+  │    └─ 强制网页模式（需该账号已在 account-manager 登录）
+  │
+  └─ target 是平台名
+       ├─ 1. 查 account-manager：有 login_status=1 的账号？ → 网页模式（免费）
+       ├─ 2. 查本地 llm_keys：有 is_active=1 的 Key？ → API Key 模式（付费）
+       └─ 3. 都没有 → 报错，给出两种凭据的配置指引
+```
+
+## 使用网页模式的前提
+
+网页模式依赖 **account-manager** 管理的账号和 Chrome Profile，使用前需完成：
+
+```bash
+# 1. 在 account-manager 中添加对应平台账号
+python3 {accountManagerDir}/scripts/main.py add "Kimi" "手机号"
+
+# 2. 登录（打开浏览器，手动完成登录后自动写入状态）
+python3 {accountManagerDir}/scripts/main.py login <id>
+```
+
+## 环境变量
+
+| 变量 | 说明 | 默认值 |
+|---|---|---|
+| `JIANGCHANG_DATA_ROOT` | 数据根目录（与 account-manager 一致） | Win: `D:\jiangchang-data` |
+| `JIANGCHANG_USER_ID` | 用户/工作区 ID | `_anon` |
+
+API Key 存储路径：`{JIANGCHANG_DATA_ROOT}/{JIANGCHANG_USER_ID}/llm-manager/llm-manager.db`  
+网页账号：通过 `account-manager` 子命令 `pick-logged-in` 查询，不直接读其数据库。
+
+## 支持的平台
+
+| 平台 | Slug | 中文别名 | 网页模式 | API 模式 |
+|---|---|---|---|---|
+| 豆包 | doubao | 豆包 | ✓ | ✓（火山方舟，需 ep-xxx） |
+| DeepSeek | deepseek | 深度求索 | ✓ | ✓ |
+| 通义千问 | qianwen | 通义、千问 | ✓ | ✓ |
+| Kimi | kimi | 月之暗面 | ✓ | ✓ |
+| 文心一言 | yiyan | 文心、一言 | ✓ | ✓ |
+| 腾讯元宝 | yuanbao | 元宝 | ✓ | ✗（暂无公开 API） |
+
+## 输出格式
+
+generate 命令输出固定格式，方便下游 Skill（如 sohu-publisher）精确提取：
+
+```
+===LLM_START===
+（生成的内容）
+===LLM_END===
+```

llm-manager/release.ps1

@@ -0,0 +1,220 @@
+<#
+.SYNOPSIS
+  One-command release script for skill repos.
+
+.DESCRIPTION
+  - Optional auto-commit
+  - Push current branch
+  - Auto-increment semantic tag (vX.Y.Z)
+  - Create & push tag
+  - Fail fast on unsafe states
+
+.EXAMPLES
+  # Safe mode (recommended): requires clean working tree
+  .\release.ps1
+
+  # Auto commit tracked/untracked changes then release
+  .\release.ps1 -AutoCommit -CommitMessage "chore: update skill config"
+
+  # Dry run (show what would happen)
+  .\release.ps1 -DryRun
+
+  # Custom tag prefix
+  .\release.ps1 -Prefix "v" -Message "正式发布"
+
+.NOTES
+  Requires: git, PowerShell 5+
+#>
+
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+if (Test-Path $sharedScript) {
+  & $sharedScript @PSBoundParameters
+  exit $LASTEXITCODE
+}
+
+function Invoke-Git {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  Write-Host ">> git $Args" -ForegroundColor DarkGray
+  & cmd /c "git $Args"
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+}
+
+function Get-GitOutput {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  $output = & cmd /c "git $Args" 2>$null
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+  return @($output)
+}
+
+function Test-Repo {
+  & git rev-parse --is-inside-work-tree *> $null
+  return ($LASTEXITCODE -eq 0)
+}
+
+function Get-CurrentBranch {
+  $b = (Get-GitOutput "branch --show-current" | Select-Object -First 1).Trim()
+  return $b
+}
+
+function Get-StatusPorcelain {
+  $lines = @(Get-GitOutput "status --porcelain")
+  return $lines
+}
+
+function Parse-SemVerTag {
+  param(
+    [string]$Tag,
+    [string]$TagPrefix
+  )
+  $escaped = [regex]::Escape($TagPrefix)
+  $m = [regex]::Match($Tag, "^${escaped}(\d+)\.(\d+)\.(\d+)$")
+  if (-not $m.Success) { return $null }
+
+  return [pscustomobject]@{
+    Raw   = $Tag
+    Major = [int]$m.Groups[1].Value
+    Minor = [int]$m.Groups[2].Value
+    Patch = [int]$m.Groups[3].Value
+  }
+}
+
+function Get-NextTag {
+  param([string]$TagPrefix)
+
+  $tags = Get-GitOutput "tag --list"
+  $parsed = @()
+
+  foreach ($t in $tags) {
+    $t = $t.Trim()
+    if (-not $t) { continue }
+    $obj = Parse-SemVerTag -Tag $t -TagPrefix $TagPrefix
+    if ($null -ne $obj) { $parsed += $obj }
+  }
+
+  if ($parsed.Count -eq 0) {
+    return "${TagPrefix}1.0.1"
+  }
+
+  $latest = $parsed | Sort-Object Major, Minor, Patch | Select-Object -Last 1
+  return "$TagPrefix$($latest.Major).$($latest.Minor).$([int]$latest.Patch + 1)"
+}
+
+function Ensure-CleanOrAutoCommit {
+  param(
+    [switch]$DoAutoCommit,
+    [switch]$NeedClean,
+    [switch]$IsDryRun,
+    [string]$Msg
+  )
+
+  $status = @(Get-StatusPorcelain)
+  if ($status.Length -eq 0) { return }
+
+  if ($NeedClean) {
+    Write-Host "Working tree is not clean and -RequireClean is enabled." -ForegroundColor Yellow
+    & git status --short
+    throw "Abort: dirty working tree."
+  }
+
+  # 默认一键发布：有改动就自动提交；也可用 -AutoCommit 显式开启
+  $commitMsg = $Msg
+  if ([string]::IsNullOrWhiteSpace($commitMsg)) {
+    $commitMsg = "chore: auto release commit ($(Get-Date -Format 'yyyy-MM-dd HH:mm:ss'))"
+  }
+
+  if (-not $DoAutoCommit) {
+    Write-Host "Detected uncommitted changes, auto-committing before release..." -ForegroundColor Yellow
+  }
+
+  if ($IsDryRun) {
+    Write-Host "[DryRun] Would run: git add -A" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git commit -m `"$commitMsg`"" -ForegroundColor Yellow
+    return
+  }
+
+  Invoke-Git "add -A"
+  Invoke-Git "commit -m `"$commitMsg`""
+}
+
+try {
+  Write-Host "=== Release Script Start ===" -ForegroundColor Cyan
+
+  if (-not (Test-Repo)) {
+    throw "Current directory is not a git repository."
+  }
+
+  $branch = Get-CurrentBranch
+  if ([string]::IsNullOrWhiteSpace($branch)) {
+    throw "Unable to determine current branch."
+  }
+
+  if ($branch -notin @("main", "master")) {
+    throw "Current branch is '$branch'. Release is only allowed from main/master."
+  }
+
+  Invoke-Git "fetch --tags --prune origin"
+
+  Ensure-CleanOrAutoCommit -DoAutoCommit:$AutoCommit -NeedClean:$RequireClean -IsDryRun:$DryRun -Msg $CommitMessage
+
+  $upstream = (& git rev-parse --abbrev-ref --symbolic-full-name "@{u}" 2>$null)
+  $hasUpstream = ($LASTEXITCODE -eq 0)
+
+  if ($DryRun) {
+    if ($hasUpstream) {
+      Write-Host "[DryRun] Would run: git push" -ForegroundColor Yellow
+    } else {
+      Write-Host "[DryRun] Would run: git push -u origin $branch" -ForegroundColor Yellow
+    }
+  } else {
+    if ($hasUpstream) {
+      Invoke-Git "push"
+    } else {
+      Invoke-Git "push -u origin $branch"
+    }
+  }
+
+  $nextTag = Get-NextTag -TagPrefix $Prefix
+  Write-Host "Next tag: $nextTag" -ForegroundColor Green
+
+  $existing = @(Get-GitOutput "tag --list `"$nextTag`"")
+  if ($existing.Length -gt 0) {
+    throw "Tag already exists: $nextTag"
+  }
+
+  if ($DryRun) {
+    Write-Host "[DryRun] Would run: git tag -a $nextTag -m `"$Message`"" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git push origin $nextTag" -ForegroundColor Yellow
+    Write-Host "=== DryRun Complete ===" -ForegroundColor Cyan
+    exit 0
+  }
+
+  Invoke-Git "tag -a $nextTag -m `"$Message`""
+  Invoke-Git "push origin $nextTag"
+
+  Write-Host "Release success: $nextTag" -ForegroundColor Green
+  Write-Host "=== Release Script Done ===" -ForegroundColor Cyan
+  exit 0
+}
+catch {
+  Write-Host "Release failed: $($_.Exception.Message)" -ForegroundColor Red
+  exit 1
+}

llm-manager/scripts/db.py

@@ -0,0 +1,292 @@
+"""
+llm-manager 本地数据库：
+- llm_keys: API Key 记录
+- llm_web_accounts: 网页模式账号关联记录（账号主数据仍由 account-manager 管理）
+"""
+import os
+import sqlite3
+import time
+from typing import Optional
+
+from providers import get_data_root, get_user_id
+
+SKILL_SLUG = "llm-manager"
+
+# SQLite 无独立 DATETIME：时间统一存 INTEGER Unix 秒（UTC）。
+LLM_KEYS_TABLE_SQL = """
+CREATE TABLE llm_keys (
+    id            INTEGER PRIMARY KEY AUTOINCREMENT,  -- 主键（自增）
+    provider      TEXT NOT NULL,        -- 平台 slug：doubao/deepseek/qianwen/kimi/yiyan/yuanbao
+    label         TEXT NOT NULL DEFAULT '',  -- 用户自定义备注，如「公司Key」
+    api_key       TEXT NOT NULL,         -- API Key 原文
+    default_model TEXT,                  -- 默认模型（doubao 须填 ep-xxx）
+    is_active     INTEGER NOT NULL DEFAULT 1,  -- 是否启用：0 停用 1 启用
+    last_used_at  INTEGER,               -- 最近调用时间，Unix 秒；从未用过为 NULL
+    created_at    INTEGER NOT NULL,
+    updated_at    INTEGER NOT NULL
+);
+"""
+
+LLM_WEB_ACCOUNTS_TABLE_SQL = """
+CREATE TABLE llm_web_accounts (
+    id            INTEGER PRIMARY KEY AUTOINCREMENT,
+    provider      TEXT NOT NULL,
+    account_id    INTEGER NOT NULL,
+    account_name  TEXT NOT NULL DEFAULT '',
+    login_status  INTEGER NOT NULL DEFAULT 1,
+    created_at    INTEGER NOT NULL,
+    updated_at    INTEGER NOT NULL,
+    UNIQUE(provider, account_id)
+);
+"""
+
+
+def _now_unix() -> int:
+    return int(time.time())
+
+
+def get_skill_data_dir() -> str:
+    path = os.path.join(get_data_root(), get_user_id(), SKILL_SLUG)
+    os.makedirs(path, exist_ok=True)
+    return path
+
+
+def get_db_path() -> str:
+    return os.path.join(get_skill_data_dir(), f"{SKILL_SLUG}.db")
+
+
+def get_conn():
+    return sqlite3.connect(get_db_path())
+
+
+def init_db():
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='llm_keys'")
+        if not cur.fetchone():
+            cur.executescript(LLM_KEYS_TABLE_SQL)
+        cur.execute("SELECT name FROM sqlite_master WHERE type='table' AND name='llm_web_accounts'")
+        if not cur.fetchone():
+            cur.executescript(LLM_WEB_ACCOUNTS_TABLE_SQL)
+        conn.commit()
+    finally:
+        conn.close()
+
+
+# ---------------------------------------------------------------------------
+# CRUD
+# ---------------------------------------------------------------------------
+
+def add_key(provider: str, api_key: str, model: Optional[str] = None, label: str = "") -> int:
+    init_db()
+    now = _now_unix()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            """
+            INSERT INTO llm_keys (provider, label, api_key, default_model, is_active, created_at, updated_at)
+            VALUES (?, ?, ?, ?, 1, ?, ?)
+            """,
+            (provider, label or "", api_key, model, now, now),
+        )
+        new_id = cur.lastrowid
+        conn.commit()
+        return new_id
+    finally:
+        conn.close()
+
+
+def upsert_web_account(provider: str, account_id: int, account_name: str = "", login_status: int = 1) -> int:
+    init_db()
+    now = _now_unix()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            """
+            INSERT INTO llm_web_accounts (provider, account_id, account_name, login_status, created_at, updated_at)
+            VALUES (?, ?, ?, ?, ?, ?)
+            ON CONFLICT(provider, account_id) DO UPDATE SET
+              account_name = excluded.account_name,
+              login_status = excluded.login_status,
+              updated_at = excluded.updated_at
+            """,
+            (provider, int(account_id), account_name or "", int(login_status or 0), now, now),
+        )
+        conn.commit()
+        cur.execute(
+            "SELECT id FROM llm_web_accounts WHERE provider = ? AND account_id = ?",
+            (provider, int(account_id)),
+        )
+        row = cur.fetchone()
+        return int(row[0]) if row else 0
+    finally:
+        conn.close()
+
+
+def list_keys(provider: Optional[str] = None, limit: int = 10) -> list:
+    init_db()
+    if not isinstance(limit, int) or limit <= 0:
+        limit = 10
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        if provider:
+            cur.execute(
+                "SELECT id, provider, label, api_key, default_model, is_active, last_used_at, created_at "
+                "FROM llm_keys WHERE provider = ? ORDER BY created_at DESC, id DESC LIMIT ?",
+                (provider, limit),
+            )
+        else:
+            cur.execute(
+                "SELECT id, provider, label, api_key, default_model, is_active, last_used_at, created_at "
+                "FROM llm_keys ORDER BY created_at DESC, id DESC LIMIT ?",
+                (limit,),
+            )
+        rows = cur.fetchall()
+    finally:
+        conn.close()
+
+    result = []
+    for row in rows:
+        result.append({
+            "id": row[0],
+            "provider": row[1],
+            "label": row[2] or "",
+            "api_key": row[3],
+            "default_model": row[4] or "",
+            "is_active": row[5],
+            "last_used_at": row[6],
+            "created_at": row[7],
+        })
+    return result
+
+
+def list_web_accounts(provider: Optional[str] = None, limit: int = 10) -> list:
+    init_db()
+    if not isinstance(limit, int) or limit <= 0:
+        limit = 10
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        if provider:
+            cur.execute(
+                "SELECT id, provider, account_id, account_name, login_status, created_at, updated_at "
+                "FROM llm_web_accounts WHERE provider = ? ORDER BY created_at DESC, id DESC LIMIT ?",
+                (provider, limit),
+            )
+        else:
+            cur.execute(
+                "SELECT id, provider, account_id, account_name, login_status, created_at, updated_at "
+                "FROM llm_web_accounts ORDER BY created_at DESC, id DESC LIMIT ?",
+                (limit,),
+            )
+        rows = cur.fetchall()
+    finally:
+        conn.close()
+
+    result = []
+    for row in rows:
+        result.append({
+            "id": row[0],
+            "provider": row[1],
+            "account_id": row[2],
+            "account_name": row[3] or "",
+            "login_status": int(row[4] or 0),
+            "created_at": row[5],
+            "updated_at": row[6],
+        })
+    return result
+
+
+def get_key_by_id(key_id: int) -> Optional[dict]:
+    init_db()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            "SELECT id, provider, label, api_key, default_model, is_active, last_used_at, created_at "
+            "FROM llm_keys WHERE id = ?",
+            (key_id,),
+        )
+        row = cur.fetchone()
+        if not row:
+            return None
+        return {
+            "id": row[0],
+            "provider": row[1],
+            "label": row[2] or "",
+            "api_key": row[3],
+            "default_model": row[4] or "",
+            "is_active": row[5],
+            "last_used_at": row[6],
+            "created_at": row[7],
+        }
+    finally:
+        conn.close()
+
+
+def delete_key(key_id: int) -> bool:
+    init_db()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute("SELECT id FROM llm_keys WHERE id = ?", (key_id,))
+        if not cur.fetchone():
+            return False
+        cur.execute("DELETE FROM llm_keys WHERE id = ?", (key_id,))
+        conn.commit()
+        return True
+    finally:
+        conn.close()
+
+
+def find_active_key(provider: str) -> Optional[dict]:
+    """查找该平台第一个 is_active=1 的 key（按 id 升序）。"""
+    init_db()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            "SELECT id, provider, label, api_key, default_model, is_active, last_used_at "
+            "FROM llm_keys WHERE provider = ? AND is_active = 1 ORDER BY id LIMIT 1",
+            (provider,),
+        )
+        row = cur.fetchone()
+        if not row:
+            return None
+        return {
+            "id": row[0],
+            "provider": row[1],
+            "label": row[2] or "",
+            "api_key": row[3],
+            "default_model": row[4] or "",
+            "is_active": row[5],
+            "last_used_at": row[6],
+        }
+    finally:
+        conn.close()
+
+
+def mark_key_used(key_id: int):
+    now = _now_unix()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            "UPDATE llm_keys SET last_used_at = ?, updated_at = ? WHERE id = ?",
+            (now, now, key_id),
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
+
+def _mask_key(api_key: str) -> str:
+    """展示时打码：前4位 + ... + 后4位。"""
+    k = api_key or ""
+    if len(k) <= 8:
+        return k[:2] + "****"
+    return k[:4] + "..." + k[-4:]

llm-manager/scripts/engines/api_engine.py

@@ -0,0 +1,28 @@
+"""
+OpenAI 兼容 API 引擎：适用于 DeepSeek、通义千问、Kimi、文心一言、豆包（火山方舟）。
+只要平台提供 OpenAI 兼容接口，均可用此引擎，无需 Playwright。
+"""
+
+
+class ApiEngine:
+    def __init__(self, api_base: str, api_key: str, model: str):
+        self.api_base = api_base
+        self.api_key = api_key
+        self.model = model
+
+    def generate(self, prompt: str) -> str:
+        try:
+            from openai import OpenAI
+        except ImportError:
+            return "ERROR:缺少依赖：pip install openai"
+
+        try:
+            client = OpenAI(base_url=self.api_base, api_key=self.api_key)
+            response = client.chat.completions.create(
+                model=self.model,
+                messages=[{"role": "user", "content": prompt}],
+            )
+            content = response.choices[0].message.content
+            return (content or "").strip()
+        except Exception as e:
+            return f"ERROR:API调用失败：{e}"

llm-manager/scripts/engines/base.py

@@ -0,0 +1,15 @@
+from abc import ABC, abstractmethod
+from playwright.async_api import Page
+
+
+class BaseEngine(ABC):
+    def __init__(self, page: Page):
+        self.page = page
+
+    @abstractmethod
+    async def generate(self, prompt: str) -> str:
+        """
+        基于操作页面的具体大模型的生成逻辑。输入文本，抓取并返回文本。
+        返回值：正常结果字符串，或以 "ERROR:" 开头的错误描述。
+        """
+        pass

llm-manager/scripts/engines/deepseek.py

@@ -0,0 +1,70 @@
+"""
+DeepSeek 网页版驱动引擎（chat.deepseek.com）。
+
+选择器说明（如页面改版需更新）：
+- 输入框：#chat-input（textarea）
+- 发送按钮：[aria-label="发送消息"] 或 div[class*="send"] > button
+- 停止生成按钮：[aria-label="停止生成"]（生成中可见，生成结束消失）
+- 复制按钮：最后一条回复下的 [aria-label="复制"]
+"""
+import asyncio
+from .base import BaseEngine
+
+
+class DeepSeekEngine(BaseEngine):
+    async def generate(self, prompt: str) -> str:
+        await self.page.goto("https://chat.deepseek.com")
+        await self.page.wait_for_load_state("networkidle")
+
+        # 登录检测：若能找到输入框则已登录
+        try:
+            editor = self.page.locator("textarea#chat-input").first
+            await editor.wait_for(state="visible", timeout=10000)
+        except Exception:
+            return "ERROR:REQUIRE_LOGIN"
+
+        # 输入提示词
+        await editor.click()
+        await self.page.keyboard.insert_text(prompt)
+        await asyncio.sleep(0.5)
+
+        # 发送（优先点按钮，失败则按 Enter）
+        sent = False
+        for sel in ('[aria-label="发送消息"]', 'div[class*="send"] > button', 'button[type="submit"]'):
+            try:
+                btn = self.page.locator(sel).first
+                if await btn.is_visible(timeout=1000):
+                    await btn.click()
+                    sent = True
+                    break
+            except Exception:
+                continue
+        if not sent:
+            await self.page.keyboard.press("Enter")
+
+        print("💡 [llm-manager/deepseek] 已发送提示词，等待 DeepSeek 生成响应...")
+        await asyncio.sleep(2)
+
+        # 等待生成完毕：停止按钮消失即为完成，超时 150 秒
+        stop_sel = '[aria-label="停止生成"]'
+        deadline = asyncio.get_event_loop().time() + 150
+        while asyncio.get_event_loop().time() < deadline:
+            try:
+                visible = await self.page.locator(stop_sel).first.is_visible(timeout=500)
+                if not visible:
+                    break
+            except Exception:
+                break
+            await asyncio.sleep(2)
+
+        await asyncio.sleep(2)
+
+        # 通过最后一个复制按钮取结果
+        try:
+            copy_btn = self.page.locator('[aria-label="复制"]').last
+            await copy_btn.click()
+            await asyncio.sleep(0.5)
+            result = await self.page.evaluate("navigator.clipboard.readText()")
+            return result.strip()
+        except Exception as e:
+            return f"ERROR:抓取 DeepSeek 内容时异常：{e}"

llm-manager/scripts/engines/doubao.py

@@ -0,0 +1,222 @@
+"""
+豆包网页版驱动（doubao.com/chat/）。
+
+生成后不依赖「喇叭瞬时出现/消失」这种不稳定信号。
+而是轮询判断两种输出是否已进入可复制状态：
+1) 右侧写作模式：出现「改用对话直接回答」入口并可点右侧复制按钮
+2) 左侧对话模式：出现可点的消息复制按钮（拿最后一条新增回复）
+"""
+
+import asyncio
+import time
+
+from .base import BaseEngine
+
+_DOUBAO_SPEAKER_PATH_SNIPPET = "M19.8628 9.29346C20.3042"
+
+
+class DoubaoEngine(BaseEngine):
+    async def generate(self, prompt: str) -> str:
+        self.page.set_default_timeout(60_000)
+        await self.page.goto(
+            "https://www.doubao.com/chat/",
+            wait_until="domcontentloaded",
+            timeout=60_000,
+        )
+
+        editor = self.page.locator('textarea[data-testid="chat_input_input"]').first
+        try:
+            await editor.wait_for(state="visible", timeout=30_000)
+        except Exception:
+            return "ERROR:REQUIRE_LOGIN"
+
+        text = (prompt or "").strip()
+        if not text:
+            return "ERROR:PROMPT_EMPTY"
+
+        await editor.click()
+        await editor.fill(text)
+        await asyncio.sleep(0.2)
+
+        send = self.page.locator(
+            '#flow-end-msg-send, [data-testid="chat_input_send_button"]'
+        ).first
+        try:
+            await send.wait_for(state="visible", timeout=15_000)
+            await send.click(timeout=10_000)
+        except Exception as e:
+            return f"ERROR:DOUBAO_SEND_FAILED {e}"
+
+        print("💡 [llm-manager/doubao] 已发送，等待生成完成后再复制（最长 180s）...")
+
+        # 发送前记录：用于确认点的是新增复制按钮，且剪贴板确实变化了
+        left_copy = self.page.locator('[data-testid="message_action_copy"]')
+        right_copy = self.page.locator('[data-testid="container_inner_copy_btn"]')
+        left_prev = await left_copy.count()
+        right_prev = await right_copy.count()
+        clipboard_before = await self._safe_read_clipboard()
+
+        # 稳定检测：避免生成中“喇叭短暂出现”误判（需要连续多次可见）
+        speaker = self.page.locator(
+            f'svg path[d*="{_DOUBAO_SPEAKER_PATH_SNIPPET}"]'
+        ).last
+        stable_needed = 3
+        stable = 0
+        interval_sec = 0.5
+        deadline = time.monotonic() + 180.0
+        while time.monotonic() < deadline:
+            try:
+                if await speaker.is_visible():
+                    stable += 1
+                else:
+                    stable = 0
+            except Exception:
+                stable = 0
+            if stable >= stable_needed:
+                break
+            await asyncio.sleep(interval_sec)
+        else:
+            return "ERROR:DOUBAO_WAIT_COMPLETE_TIMEOUT 180 秒内未检测到“喇叭稳定回显”。"
+
+        # 生成完成后：依据当前页面状态决定点右侧还是左侧复制
+        write_mode = await self._is_write_mode()
+        if write_mode:
+            if await self._wait_count_gt(right_copy, right_prev, timeout_sec=15.0):
+                copy_btn = right_copy.last
+            elif await self._wait_count_gt(left_copy, left_prev, timeout_sec=15.0):
+                copy_btn = left_copy.last
+            else:
+                return "ERROR:DOUBAO_COPY_BUTTON_NOT_READY 右侧/左侧复制按钮均未出现新增。"
+        else:
+            if await self._wait_count_gt(left_copy, left_prev, timeout_sec=15.0):
+                copy_btn = left_copy.last
+            elif await self._wait_count_gt(right_copy, right_prev, timeout_sec=5.0):
+                copy_btn = right_copy.last
+            else:
+                return "ERROR:DOUBAO_COPY_BUTTON_NOT_READY 左侧复制按钮未出现新增。"
+
+        try:
+            await copy_btn.scroll_into_view_if_needed()
+            await copy_btn.click(timeout=15_000, force=True)
+        except Exception as e:
+            return f"ERROR:DOUBAO_COPY_CLICK_FAILED {e}"
+
+        out = await self._wait_clipboard_changed(clipboard_before, timeout_sec=20.0)
+        if out is None:
+            return "ERROR:DOUBAO_CLIPBOARD_NOT_UPDATED 剪贴板未在规定时间内更新。"
+        if self._is_invalid_clipboard(out, clipboard_before=clipboard_before):
+            return f"ERROR:DOUBAO_CLIPBOARD_INVALID {out[:200]}"
+
+        await asyncio.sleep(5)
+        return out
+
+    async def _is_write_mode(self) -> bool:
+        marker = self.page.locator('div.bottom-entry-GSWErB:has-text("改用对话直接回答")').first
+        try:
+            return await marker.is_visible()
+        except Exception:
+            return False
+
+    async def _wait_count_gt(self, locator, prev_count: int, timeout_sec: float) -> bool:
+        deadline = time.monotonic() + timeout_sec
+        while time.monotonic() < deadline:
+            try:
+                if await locator.count() > prev_count:
+                    return True
+            except Exception:
+                pass
+            await asyncio.sleep(0.3)
+        return False
+
+    async def _wait_clipboard_changed(
+        self, clipboard_before: str | None, timeout_sec: float
+    ) -> str | None:
+        deadline = time.monotonic() + timeout_sec
+        last = clipboard_before
+        while time.monotonic() < deadline:
+            cur = await self._safe_read_clipboard()
+            if cur and cur != last and not self._is_invalid_clipboard(
+                cur, clipboard_before=clipboard_before
+            ):
+                return cur.strip()
+            last = cur
+            await asyncio.sleep(0.4)
+        return None
+
+    async def _try_copy_right(self, right_copy) -> str | None:
+        # 右侧复制按钮
+        try:
+            btn = right_copy.first
+            await btn.wait_for(state="visible", timeout=2000)
+            await btn.scroll_into_view_if_needed()
+            await btn.click(timeout=5000, force=True)
+        except Exception:
+            return None
+
+        await asyncio.sleep(0.4)
+        out = await self._safe_read_clipboard()
+        if self._is_invalid_clipboard(out):
+            return None
+        if out == (await self._safe_read_clipboard()):
+            pass
+        return (out or "").strip()
+
+    async def _try_copy_left(self, left_copy, prev_count: int) -> str | None:
+        n = await left_copy.count()
+        if n <= prev_count:
+            return None
+
+        # 从新增区间末尾往前找第一个可点击的复制按钮
+        for idx in range(n - 1, prev_count - 1, -1):
+            btn = left_copy.nth(idx)
+            try:
+                if not await btn.is_visible():
+                    continue
+                await btn.scroll_into_view_if_needed()
+                await btn.click(timeout=5000, force=True)
+            except Exception:
+                continue
+
+            await asyncio.sleep(0.4)
+            out = await self._safe_read_clipboard()
+            if self._is_invalid_clipboard(out):
+                continue
+            return (out or "").strip()
+        return None
+
+    def _is_invalid_clipboard(
+        self, text: str | None, clipboard_before: str | None = None
+    ) -> bool:
+        if not text:
+            return True
+        t = str(text).strip()
+        if not t:
+            return True
+        if clipboard_before is not None and t == str(clipboard_before).strip():
+            return True
+        if t.startswith("ERROR:"):
+            return True
+        # 防止误点拿到 UI 文案/占位文本
+        banned_substrings = (
+            "改用对话直接回答",
+            "新对话",
+            "正在生成",
+            "复制",
+            "下载",
+            "快捷",
+            "发消息",
+        )
+        if any(s in t for s in banned_substrings):
+            return True
+        if "data-testid" in t or "<button" in t:
+            return True
+        # 经验阈值：正常正文不会太短
+        if len(t) < 30:
+            return True
+        return False
+
+    async def _safe_read_clipboard(self) -> str | None:
+        try:
+            return await self.page.evaluate("navigator.clipboard.readText()")
+        except Exception:
+            return None

llm-manager/scripts/engines/kimi.py

@@ -0,0 +1,48 @@
+import asyncio
+from .base import BaseEngine
+
+
+class KimiEngine(BaseEngine):
+    async def generate(self, prompt: str) -> str:
+        await self.page.goto("https://kimi.moonshot.cn/")
+        await self.page.wait_for_load_state("networkidle")
+
+        # 登录检测：等待输入框出现，超时则未登录
+        try:
+            editor = self.page.locator(".chat-input-editor").first
+            await editor.wait_for(state="visible", timeout=10000)
+        except Exception:
+            return "ERROR:REQUIRE_LOGIN"
+
+        # 输入提示词（insert_text 避免换行符被误触发）
+        await editor.click()
+        await self.page.keyboard.insert_text(prompt)
+        await asyncio.sleep(0.5)
+
+        # 点击发送按钮
+        await self.page.locator(".send-button-container").first.click()
+
+        print("💡 [llm-manager/kimi] 已发送提示词，等待 Kimi 生成响应...")
+
+        # 等待 2 秒让发送按钮进入"生成中"状态
+        await asyncio.sleep(2)
+
+        # 等待 Send SVG 图标重新出现（表示生成完毕）
+        try:
+            send_icon = self.page.locator('.send-button-container svg[name="Send"]')
+            await send_icon.wait_for(state="visible", timeout=150000)
+        except Exception:
+            return "ERROR:Kimi 生成超时（150秒未见完成标志），可能网络卡住或内容被拦截。"
+
+        # 稳妥起见多等 2 秒，让复制按钮完全渲染
+        await asyncio.sleep(2)
+
+        # 点击最后一条回复的复制按钮，通过剪贴板取完整文本
+        try:
+            copy_btn = self.page.locator('svg[name="Copy"]').last
+            await copy_btn.click()
+            await asyncio.sleep(0.5)
+            result = await self.page.evaluate("navigator.clipboard.readText()")
+            return result.strip()
+        except Exception as e:
+            return f"ERROR:抓取 Kimi 内容时异常：{e}"

llm-manager/scripts/engines/qianwen.py

@@ -0,0 +1,91 @@
+"""
+通义千问网页版驱动引擎（tongyi.aliyun.com/qianwen）。
+
+选择器说明（如页面改版需更新）：
+- 输入框：#search-input 或 textarea[class*="input"]（textarea）
+- 发送按钮：button[class*="send"] 或 [aria-label="发送"]
+- 停止生成：button[class*="stop"] 或 [aria-label="停止"]
+- 复制按钮：最后一条回复下的 [class*="copy"] 或 [aria-label="复制"]
+"""
+import asyncio
+from .base import BaseEngine
+
+
+class QianwenEngine(BaseEngine):
+    async def generate(self, prompt: str) -> str:
+        await self.page.goto("https://tongyi.aliyun.com/qianwen/")
+        await self.page.wait_for_load_state("networkidle")
+
+        # 登录检测
+        input_selectors = [
+            "#search-input",
+            "textarea[class*='input']",
+            "div[class*='input'][contenteditable='true']",
+        ]
+        editor = None
+        for sel in input_selectors:
+            try:
+                loc = self.page.locator(sel).first
+                await loc.wait_for(state="visible", timeout=4000)
+                editor = loc
+                break
+            except Exception:
+                continue
+        if editor is None:
+            return "ERROR:REQUIRE_LOGIN"
+
+        # 输入提示词
+        await editor.click()
+        await self.page.keyboard.insert_text(prompt)
+        await asyncio.sleep(0.5)
+
+        # 发送
+        sent = False
+        for sel in (
+            'button[class*="send"]',
+            '[aria-label="发送"]',
+            'button[type="submit"]',
+        ):
+            try:
+                btn = self.page.locator(sel).first
+                if await btn.is_visible(timeout=1000):
+                    await btn.click()
+                    sent = True
+                    break
+            except Exception:
+                continue
+        if not sent:
+            await self.page.keyboard.press("Enter")
+
+        print("💡 [llm-manager/qianwen] 已发送提示词，等待通义千问生成响应...")
+        await asyncio.sleep(2)
+
+        # 等待生成完毕
+        stop_selectors = ['button[class*="stop"]', '[aria-label="停止"]', '[aria-label="停止生成"]']
+        deadline = asyncio.get_event_loop().time() + 150
+        while asyncio.get_event_loop().time() < deadline:
+            stop_visible = False
+            for sel in stop_selectors:
+                try:
+                    if await self.page.locator(sel).first.is_visible(timeout=300):
+                        stop_visible = True
+                        break
+                except Exception:
+                    pass
+            if not stop_visible:
+                break
+            await asyncio.sleep(2)
+
+        await asyncio.sleep(2)
+
+        # 取结果
+        try:
+            copy_btn = self.page.locator(
+                '[aria-label="复制"], [class*="copy-btn"], button:has(svg[class*="copy"])'
+            ).last
+            await copy_btn.click()
+            await asyncio.sleep(0.5)
+            result = await self.page.evaluate("navigator.clipboard.readText()")
+            return result.strip()
+        except Exception as e:
+            return f"ERROR:抓取通义千问内容时异常：{e}"

llm-manager/scripts/engines/yiyan.py

@@ -0,0 +1,91 @@
+"""
+文心一言网页版驱动引擎（yiyan.baidu.com）。
+
+选择器说明（如页面改版需更新）：
+- 输入框：[class*="editor"] 或 div[contenteditable="true"]（富文本编辑器）
+- 发送按钮：[class*="send-btn"] 或 [aria-label="发送"]
+- 停止生成：[class*="stop"] 相关按钮
+- 复制按钮：最后一条回复下的复制按钮
+"""
+import asyncio
+from .base import BaseEngine
+
+
+class YiyanEngine(BaseEngine):
+    async def generate(self, prompt: str) -> str:
+        await self.page.goto("https://yiyan.baidu.com")
+        await self.page.wait_for_load_state("networkidle")
+
+        # 登录检测
+        input_selectors = [
+            "div[class*='editor'][contenteditable='true']",
+            "textarea[class*='input']",
+            "[contenteditable='true']",
+        ]
+        editor = None
+        for sel in input_selectors:
+            try:
+                loc = self.page.locator(sel).first
+                await loc.wait_for(state="visible", timeout=4000)
+                editor = loc
+                break
+            except Exception:
+                continue
+        if editor is None:
+            return "ERROR:REQUIRE_LOGIN"
+
+        # 输入提示词
+        await editor.click()
+        await self.page.keyboard.insert_text(prompt)
+        await asyncio.sleep(0.5)
+
+        # 发送
+        sent = False
+        for sel in (
+            '[class*="send-btn"]',
+            '[aria-label="发送"]',
+            'button[class*="send"]',
+        ):
+            try:
+                btn = self.page.locator(sel).first
+                if await btn.is_visible(timeout=1000):
+                    await btn.click()
+                    sent = True
+                    break
+            except Exception:
+                continue
+        if not sent:
+            await self.page.keyboard.press("Enter")
+
+        print("💡 [llm-manager/yiyan] 已发送提示词，等待文心一言生成响应...")
+        await asyncio.sleep(2)
+
+        # 等待生成完毕
+        stop_selectors = ['[class*="stop"]', '[aria-label="停止"]', '[aria-label="停止生成"]']
+        deadline = asyncio.get_event_loop().time() + 150
+        while asyncio.get_event_loop().time() < deadline:
+            stop_visible = False
+            for sel in stop_selectors:
+                try:
+                    if await self.page.locator(sel).first.is_visible(timeout=300):
+                        stop_visible = True
+                        break
+                except Exception:
+                    pass
+            if not stop_visible:
+                break
+            await asyncio.sleep(2)
+
+        await asyncio.sleep(2)
+
+        # 取结果
+        try:
+            copy_btn = self.page.locator(
+                '[aria-label="复制"], [class*="copy"], button:has(svg[class*="copy"])'
+            ).last
+            await copy_btn.click()
+            await asyncio.sleep(0.5)
+            result = await self.page.evaluate("navigator.clipboard.readText()")
+            return result.strip()
+        except Exception as e:
+            return f"ERROR:抓取文心一言内容时异常：{e}"

llm-manager/scripts/engines/yuanbao.py

@@ -0,0 +1,94 @@
+"""
+腾讯元宝网页版驱动引擎（yuanbao.tencent.com）。
+元宝暂无公开 API，仅支持网页模式。
+
+选择器说明（如页面改版需更新）：
+- 输入框：[class*="input-area"] textarea 或 [contenteditable="true"]
+- 发送按钮：[class*="send-btn"] 或 [aria-label="发送"]
+- 停止生成：[class*="stop"] 相关按钮
+- 复制按钮：最后一条回复下的复制按钮
+"""
+import asyncio
+from .base import BaseEngine
+
+
+class YuanbaoEngine(BaseEngine):
+    async def generate(self, prompt: str) -> str:
+        await self.page.goto("https://yuanbao.tencent.com/chat")
+        await self.page.wait_for_load_state("networkidle")
+
+        # 登录检测
+        input_selectors = [
+            "textarea[class*='input']",
+            "div[class*='input-area'] textarea",
+            "[contenteditable='true']",
+            "textarea",
+        ]
+        editor = None
+        for sel in input_selectors:
+            try:
+                loc = self.page.locator(sel).first
+                await loc.wait_for(state="visible", timeout=4000)
+                editor = loc
+                break
+            except Exception:
+                continue
+        if editor is None:
+            return "ERROR:REQUIRE_LOGIN"
+
+        # 输入提示词
+        await editor.click()
+        await self.page.keyboard.insert_text(prompt)
+        await asyncio.sleep(0.5)
+
+        # 发送
+        sent = False
+        for sel in (
+            '[class*="send-btn"]',
+            '[aria-label="发送"]',
+            'button[class*="send"]',
+            'button[type="submit"]',
+        ):
+            try:
+                btn = self.page.locator(sel).first
+                if await btn.is_visible(timeout=1000):
+                    await btn.click()
+                    sent = True
+                    break
+            except Exception:
+                continue
+        if not sent:
+            await self.page.keyboard.press("Enter")
+
+        print("💡 [llm-manager/yuanbao] 已发送提示词，等待腾讯元宝生成响应...")
+        await asyncio.sleep(2)
+
+        # 等待生成完毕
+        stop_selectors = ['[class*="stop"]', '[aria-label="停止"]', '[aria-label="停止生成"]']
+        deadline = asyncio.get_event_loop().time() + 150
+        while asyncio.get_event_loop().time() < deadline:
+            stop_visible = False
+            for sel in stop_selectors:
+                try:
+                    if await self.page.locator(sel).first.is_visible(timeout=300):
+                        stop_visible = True
+                        break
+                except Exception:
+                    pass
+            if not stop_visible:
+                break
+            await asyncio.sleep(2)
+
+        await asyncio.sleep(2)
+
+        # 取结果
+        try:
+            copy_btn = self.page.locator(
+                '[aria-label="复制"], [class*="copy"], button:has(svg[class*="copy"])'
+            ).last
+            await copy_btn.click()
+            await asyncio.sleep(0.5)
+            result = await self.page.evaluate("navigator.clipboard.readText()")
+            return result.strip()
+        except Exception as e:
+            return f"ERROR:抓取腾讯元宝内容时异常：{e}"

llm-manager/scripts/main.py

@@ -0,0 +1,575 @@
+"""
+llm-manager 主入口 CLI。
+
+子命令：
+  health                              快速离线健康检查
+  version                             输出版本 JSON
+  add <platform> [api_key]            添加 API Key（不传 api_key 则走网页账号关联）
+  key list [platform]                 列出 Key（打码）
+  key del <key_id>                    删除 Key
+  generate <platform_or_account_id>   生成内容（优先网页模式，备用 API Key 模式）
+            "<prompt>"
+
+generate 调度规则：
+  1. 若 target 为纯数字 → 视为 account-manager 账号 ID → 强制网页模式
+  2. 若 target 为平台名/别名：
+     a. 先查 account-manager 有无该平台已登录账号 → 网页模式（免费）
+     b. 再查 llm_keys 有无可用 Key → API Key 模式（付费）
+     c. 两者均无 → 报错并给出操作指引
+"""
+import sys
+import json
+import os
+import asyncio
+import subprocess
+
+# Windows GBK 编码兼容修复
+if sys.platform == "win32":
+    import io
+    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
+    sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace")
+
+BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+OPENCLAW_DIR = os.path.dirname(BASE_DIR)
+
+# 确保 scripts 目录在 sys.path，使 providers/db 可直接 import
+_scripts_dir = os.path.dirname(os.path.abspath(__file__))
+if _scripts_dir not in sys.path:
+    sys.path.insert(0, _scripts_dir)
+
+from providers import (
+    LLM_PROVIDERS,
+    resolve_provider_key,
+    provider_list_cn,
+    find_logged_in_account,
+    resolve_chromium_channel,
+)
+from db import (
+    add_key,
+    upsert_web_account,
+    list_keys,
+    list_web_accounts,
+    get_key_by_id,
+    delete_key,
+    find_active_key,
+    mark_key_used,
+    _mask_key,
+)
+from engines.api_engine import ApiEngine
+from engines.kimi import KimiEngine
+from engines.doubao import DoubaoEngine
+from engines.deepseek import DeepSeekEngine
+from engines.qianwen import QianwenEngine
+from engines.yiyan import YiyanEngine
+from engines.yuanbao import YuanbaoEngine
+
+# 平台 slug → 网页引擎类（全部 6 个平台）
+WEB_ENGINES = {
+    "doubao": DoubaoEngine,
+    "deepseek": DeepSeekEngine,
+    "qianwen": QianwenEngine,
+    "kimi": KimiEngine,
+    "yiyan": YiyanEngine,
+    "yuanbao": YuanbaoEngine,
+}
+
+SKILL_VERSION = "1.0.3"
+
+
+def _engine_result_is_error(text: str) -> bool:
+    """网页/API 引擎约定：失败时返回以 ERROR: 开头的字符串，不得当作正文包进 LLM 标记块。"""
+    return (text or "").lstrip().startswith("ERROR:")
+
+
+def _unix_to_iso(ts):
+    if ts is None:
+        return ""
+    try:
+        import datetime as _dt
+
+        return _dt.datetime.fromtimestamp(int(ts)).isoformat(timespec="seconds")
+    except Exception:
+        return str(ts)
+
+
+# ---------------------------------------------------------------------------
+# account-manager 跨 Skill 调用
+# ---------------------------------------------------------------------------
+
+def _get_account_from_manager(account_id) -> dict | None:
+    """通过 account-manager CLI 按 ID 取账号 JSON。"""
+    script = os.path.join(OPENCLAW_DIR, "account-manager", "scripts", "main.py")
+    try:
+        result = subprocess.run(
+            [sys.executable, script, "get", str(account_id)],
+            capture_output=True, text=True, encoding="utf-8", errors="replace",
+        )
+        raw = result.stdout.strip()
+        if raw.startswith("ERROR"):
+            return None
+        return json.loads(raw)
+    except Exception:
+        return None
+
+
+# ---------------------------------------------------------------------------
+# 网页模式
+# ---------------------------------------------------------------------------
+
+async def _run_web_generate(account: dict, prompt: str) -> bool:
+    from playwright.async_api import async_playwright
+
+    platform = account.get("platform", "")
+    profile_dir = account.get("profile_dir", "").strip()
+
+    engine_cls = WEB_ENGINES.get(platform)
+    if not engine_cls:
+        print(f"ERROR:ENGINE_NOT_FOUND 平台 {platform} 暂无网页引擎实现。")
+        return False
+
+    if not profile_dir or not os.path.isdir(profile_dir):
+        print(f"ERROR:PROFILE_DIR_MISSING 账号 {account.get('id')} 的浏览器数据目录不存在：{profile_dir}")
+        print("请先通过 account-manager 执行登录：python account-manager/scripts/main.py login <id>")
+        return False
+
+    channel = resolve_chromium_channel()
+    if not channel:
+        print("ERROR:浏览器未找到。请安装 Google Chrome 或 Microsoft Edge 后重试。")
+        return False
+
+    print(f"[llm-manager] 使用网页模式 | 平台: {LLM_PROVIDERS.get(platform, {}).get('label', platform)} | 账号: {account.get('name', account.get('id'))}")
+
+    async with async_playwright() as p:
+        browser = await p.chromium.launch_persistent_context(
+            user_data_dir=profile_dir,
+            headless=False,
+            channel=channel,
+            no_viewport=True,
+            permissions=["clipboard-read", "clipboard-write"],
+            args=["--start-maximized"],
+        )
+        try:
+            page = browser.pages[0] if browser.pages else await browser.new_page()
+            engine = engine_cls(page)
+            try:
+                result = await engine.generate(prompt)
+            except Exception as e:
+                print(f"ERROR:WEB_GENERATE_FAILED 网页模式生成失败：{e}")
+                return False
+
+            result = (result or "").strip()
+            if _engine_result_is_error(result):
+                print(result)
+                return False
+
+            print("===LLM_START===")
+            print(result)
+            print("===LLM_END===")
+            return True
+        finally:
+            await asyncio.sleep(3)
+            await browser.close()
+
+
+# ---------------------------------------------------------------------------
+# API Key 模式
+# ---------------------------------------------------------------------------
+
+def _run_api_generate(provider_key: str, key_record: dict, prompt: str) -> bool:
+    provider = LLM_PROVIDERS[provider_key]
+    api_base = provider.get("api_base")
+    model = (key_record.get("default_model") or "").strip() or (provider.get("api_model") or "").strip()
+
+    if not api_base:
+        print(f"ERROR:NO_API_BASE {provider['label']} 暂无 API 地址（不支持 API 模式）。")
+        return False
+    if not model:
+        print(
+            f"ERROR:API_MODEL_MISSING {provider['label']} 需要指定模型名称。\n"
+            f"提示：{provider.get('api_note', '')}\n"
+            f"请删除该 Key 并重新添加时带上 --model 参数：python main.py add {provider_key} \"Key\" --model \"模型名\""
+        )
+        return False
+
+    print(f"[llm-manager] 使用 API Key 模式 | 平台: {provider['label']} | 模型: {model}")
+
+    try:
+        engine = ApiEngine(api_base=api_base, api_key=key_record["api_key"], model=model)
+        result = engine.generate(prompt)
+    except Exception as e:
+        print(f"ERROR:API_GENERATE_FAILED API 模式调用失败：{e}")
+        return False
+    result = (result or "").strip()
+    if _engine_result_is_error(result):
+        print(result)
+        return False
+    mark_key_used(key_record["id"])
+
+    print("===LLM_START===")
+    print(result)
+    print("===LLM_END===")
+    return True
+
+
+# ---------------------------------------------------------------------------
+# generate 命令
+# ---------------------------------------------------------------------------
+
+def cmd_generate(target: str, prompt: str) -> bool:
+    """
+    target 可以是：
+      - 纯数字 → account-manager 账号 ID（强制网页模式）
+      - 平台名/别名 → 自动选择模式（网页优先，API Key 备用）
+    成功返回 True，任一步失败返回 False（调用方应设进程退出码非 0）。
+    """
+    target = (target or "").strip()
+    prompt = (prompt or "").strip()
+    if not prompt:
+        print("ERROR:PROMPT_EMPTY 提示词不能为空。")
+        return False
+
+    # ---- 情况 1：显式 account_id（纯数字）→ 强制网页模式 ----
+    if target.isdigit():
+        account = _get_account_from_manager(target)
+        if not account:
+            print(f"ERROR:ACCOUNT_NOT_FOUND 未在「模型管理」对应的账号列表中找到 id={target}。")
+            print("请先在模型管理中添加该平台账号，或检查 id 是否抄错。")
+            return False
+        if account.get("login_status") != 1:
+            print(
+                f"ERROR:REQUIRE_LOGIN 账号「{account.get('name', target)}」尚未完成网页登录。\n"
+                "请先在「模型管理」里确认已添加该账号，再执行下面命令完成登录：\n"
+                f"  python account-manager/scripts/main.py login {target}\n"
+                "若使用网页模式：请先在对应平台官网注册好账号，再回到本工具添加并登录。"
+            )
+            return False
+        return asyncio.run(_run_web_generate(account, prompt))
+
+    # ---- 情况 2：平台名 → 自动选模式 ----
+    provider_key = resolve_provider_key(target)
+    if not provider_key:
+        print(f"ERROR:INVALID_PLATFORM 无法识别的平台「{target}」。")
+        print("支持的平台：" + provider_list_cn())
+        return False
+
+    provider = LLM_PROVIDERS[provider_key]
+    label = provider["label"]
+    web_url = (provider.get("web_url") or "").strip()
+
+    # 优先级 1：查 account-manager 已登录账号（网页模式，免费）
+    account = find_logged_in_account(provider_key)
+    if account:
+        return asyncio.run(_run_web_generate(account, prompt))
+
+    # 优先级 2：查本地 API Key（付费）
+    key_record = find_active_key(provider_key)
+    if key_record:
+        return _run_api_generate(provider_key, key_record, prompt)
+
+    # 两种凭据均无 → 明确说明「模型管理 + 网页模式需官网账号」
+    print(f"ERROR:NO_CREDENTIAL 当前没有可用的「{label}」调用凭据（既没有已登录的网页账号，也没有可用的 API Key）。")
+    print()
+    print("【推荐 · 网页模式（免费）】按顺序做：")
+    print(f"  ① 打开「模型管理」（与 OpenClaw 里 account-manager 账号管理是同一套数据），先把「{label}」账号添加进去。")
+    print("     命令行示例（在 OpenClaw 根目录执行）：")
+    print(f"       python account-manager/scripts/main.py add \"{label}\" \"你的手机号或登录名\"")
+    print("       python account-manager/scripts/main.py login <上一步返回的账号 id>")
+    print("  ② 若走网页模式：请先在对应平台官方网站注册并创建好账号（否则浏览器里登录会失败）。")
+    if web_url:
+        print(f"     「{label}」网页入口：{web_url}")
+    print()
+    if provider.get("has_api"):
+        print("【备选 · API Key（付费）】若已在厂商控制台开通 API，可本地登记 Key 后走接口调用：")
+        print(f"  python llm-manager/scripts/main.py add {provider_key} \"你的API Key\"")
+        if provider.get("api_note"):
+            print(f"  说明：{provider['api_note']}")
+    else:
+        print(f"说明：「{label}」暂无公开 API，只能使用上面的网页模式。")
+    return False
+
+
+# ---------------------------------------------------------------------------
+# key 子命令
+# ---------------------------------------------------------------------------
+
+def cmd_key_add(provider_input: str, api_key: str = "", model: str = None, label: str = ""):
+    provider_key = resolve_provider_key(provider_input)
+    if not provider_key:
+        print(f"ERROR:INVALID_PLATFORM 无法识别的平台「{provider_input}」。")
+        print("支持：" + provider_list_cn())
+        return
+    provider = LLM_PROVIDERS[provider_key]
+    api_key = (api_key or "").strip()
+
+    # 不传 api_key：默认走网页版本，检查 account-manager 是否已有该平台已登录账号
+    if not api_key:
+        account = find_logged_in_account(provider_key)
+        if account:
+            web_row_id = upsert_web_account(
+                provider=provider_key,
+                account_id=int(account.get("id")),
+                account_name=(account.get("name") or account.get("account_name") or ""),
+                login_status=int(account.get("login_status") or 0),
+            )
+            print(
+                f"OK:WEB_ACCOUNT_READY 已关联网页模式账号 | 平台: {provider['label']} "
+                f"| account_id: {account.get('id')} | 账号: {account.get('name') or account.get('account_name') or ''}"
+            )
+            print(f"已写入 llm-manager 记录：WEB_ID {web_row_id}")
+            print(f"后续可直接调用：python main.py generate {provider_key} \"<提示词>\"")
+            return
+        print(f"ERROR:WEB_ACCOUNT_NOT_FOUND 未找到已登录的「{provider['label']}」账号，暂时无法走网页模式。")
+        print("请先在 account-manager 添加并登录该平台账号：")
+        print(f"  python account-manager/scripts/main.py add \"{provider['label']}\" \"你的登录名\"")
+        print("  python account-manager/scripts/main.py login <账号id>")
+        if provider.get("has_api"):
+            print("或直接提供 API Key：")
+            print(f"  python main.py add {provider_key} \"<API_Key>\"")
+        return
+
+    if not provider.get("has_api"):
+        print(f"ERROR:{provider['label']} 暂无公开 API，不支持 API Key 模式。")
+        print("请改用网页模式并先在 account-manager 完成登录。")
+        return
+
+    new_id = add_key(provider=provider_key, api_key=api_key, model=model, label=label)
+    print(f"✅ 已保存 API Key：ID {new_id} | {provider['label']} | 模型: {model or provider.get('api_model') or '(未指定)'} | {_mask_key(api_key)}")
+    if not model and not provider.get("api_model"):
+        print(f"⚠️ 注意：该平台需要指定模型名称，否则调用时会报错。")
+        print(f"   {provider.get('api_note', '')}")
+        print(f"   可删除后重新添加：python main.py key del {new_id}")
+
+
+def cmd_key_list(provider_input: str = None, limit: int = 10):
+    provider_key = None
+    if provider_input:
+        provider_key = resolve_provider_key(provider_input)
+        if not provider_key:
+            print(f"ERROR:INVALID_PLATFORM 无法识别的平台「{provider_input}」。")
+            return
+
+    keys = list_keys(provider_key, limit=limit)
+    web_accounts = list_web_accounts(provider_key, limit=limit)
+    if not keys and not web_accounts:
+        print("暂无记录（API Key / 网页账号关联 都为空）。")
+        print("添加 API Key：python main.py add <platform> \"API Key\" [--model 模型名]")
+        print("添加网页账号关联：python main.py add <platform>")
+        return
+
+    sep_line = "_" * 39
+    rows = []
+    for k in keys:
+        rows.append({
+            "type": "api_key",
+            "created_at": int(k.get("created_at") or 0),
+            "payload": k,
+        })
+    for w in web_accounts:
+        rows.append({
+            "type": "web_account",
+            "created_at": int(w.get("created_at") or 0),
+            "payload": w,
+        })
+    rows.sort(key=lambda x: (x["created_at"], int(x["payload"].get("id") or 0)), reverse=True)
+
+    for idx, row in enumerate(rows):
+        if row["type"] == "api_key":
+            k = row["payload"]
+            print("record_type：api_key")
+            print(f"id：{k['id']}")
+            print(f"platform：{k['provider']}")
+            print(f"platform_cn：{LLM_PROVIDERS.get(k['provider'], {}).get('label', k['provider'])}")
+            print(f"label：{k.get('label') or ''}")
+            print(f"api_key：{_mask_key(k.get('api_key') or '')}")
+            print(f"default_model：{k.get('default_model') or ''}")
+            print(f"is_active：{int(k.get('is_active') or 0)}")
+            print(f"last_used_at：{_unix_to_iso(k.get('last_used_at'))}")
+            print(f"created_at：{_unix_to_iso(k.get('created_at'))}")
+        else:
+            w = row["payload"]
+            print("record_type：web_account")
+            print(f"id：{w['id']}")
+            print(f"platform：{w['provider']}")
+            print(f"platform_cn：{LLM_PROVIDERS.get(w['provider'], {}).get('label', w['provider'])}")
+            print(f"account_id：{w.get('account_id')}")
+            print(f"account_name：{w.get('account_name') or ''}")
+            print(f"login_status：{int(w.get('login_status') or 0)}")
+            print(f"created_at：{_unix_to_iso(w.get('created_at'))}")
+            print(f"updated_at：{_unix_to_iso(w.get('updated_at'))}")
+        if idx != len(rows) - 1:
+            print(sep_line)
+            print()
+
+
+def cmd_key_del(key_id_str: str):
+    if not key_id_str.isdigit():
+        print(f"ERROR:INVALID_ID key_id 须为正整数，收到「{key_id_str}」。")
+        return
+    key_id = int(key_id_str)
+    key = get_key_by_id(key_id)
+    if not key:
+        print(f"ERROR:KEY_NOT_FOUND 未找到 ID={key_id} 的 API Key。")
+        return
+    delete_key(key_id)
+    label_str = f"（{key['label']}）" if key.get("label") else ""
+    print(f"✅ 已删除：ID {key_id} | {LLM_PROVIDERS.get(key['provider'], {}).get('label', key['provider'])}{label_str} | {_mask_key(key['api_key'])}")
+
+
+# ---------------------------------------------------------------------------
+# health / version
+# ---------------------------------------------------------------------------
+
+def cmd_health():
+    ok = sys.version_info >= (3, 10)
+    sys.exit(0 if ok else 1)
+
+
+def cmd_version():
+    print(json.dumps({"version": SKILL_VERSION, "skill": "llm-manager"}, ensure_ascii=False))
+
+
+# ---------------------------------------------------------------------------
+# CLI 解析
+# ---------------------------------------------------------------------------
+
+def _print_usage():
+    print("用法：")
+    print("  python main.py health")
+    print("  python main.py version")
+    print("  python main.py generate <平台名或account_id> \"<提示词>\"")
+    print("  python main.py add <平台> [API_Key] [--model 模型名] [--label 备注]")
+    print("  python main.py list [平台] [--limit 条数]")
+    print("  python main.py del <key_id>")
+    print()
+    print("支持的平台：" + provider_list_cn())
+    print()
+    print("generate 说明：")
+    print("  · 传入 account_id（数字）→ 指定账号网页模式（需先用 account-manager 登录）")
+    print("  · 传入平台名 → 自动选择：优先网页模式（免费），无可用账号时才用 API Key（付费）")
+    print()
+    print("兼容说明：key add/list/del 旧写法仍可用。")
+
+
+def main(argv=None) -> int:
+    args = argv if argv is not None else sys.argv[1:]
+    if not args:
+        _print_usage()
+        return 1
+    if args[0] in ("-h", "--help"):
+        _print_usage()
+        return 0
+
+    def _parse_list_args(rest_args):
+        platform_filter = None
+        limit = 10
+        i = 0
+        while i < len(rest_args):
+            if rest_args[i] == "--limit" and i + 1 < len(rest_args):
+                try:
+                    limit = int(rest_args[i + 1])
+                except ValueError:
+                    print(f"ERROR:CLI_KEY_LIST_BAD_LIMIT limit 必须是整数，收到「{rest_args[i + 1]}」。")
+                    return None, None, 1
+                i += 2
+            else:
+                if platform_filter is None:
+                    platform_filter = rest_args[i]
+                i += 1
+        return platform_filter, limit, 0
+
+    def _parse_add_args(rest_args):
+        if len(rest_args) < 1:
+            print("ERROR:CLI_ADD_MISSING_ARGS")
+            print("用法：python main.py add <平台> [API_Key] [--model 模型] [--label 备注]")
+            return None, None, None, None, 1
+        platform_arg = rest_args[0]
+        api_key_arg = ""
+        model_arg = None
+        label_arg = ""
+        i = 1
+        if i < len(rest_args) and not rest_args[i].startswith("--"):
+            api_key_arg = rest_args[i]
+            i += 1
+        while i < len(rest_args):
+            if rest_args[i] == "--model" and i + 1 < len(rest_args):
+                model_arg = rest_args[i + 1]
+                i += 2
+            elif rest_args[i] == "--label" and i + 1 < len(rest_args):
+                label_arg = rest_args[i + 1]
+                i += 2
+            else:
+                i += 1
+        return platform_arg, api_key_arg, model_arg, label_arg, 0
+
+    cmd = args[0]
+
+    if cmd == "health":
+        cmd_health()
+
+    elif cmd == "version":
+        cmd_version()
+
+    elif cmd == "generate":
+        if len(args) < 3:
+            print("ERROR:CLI_GENERATE_MISSING_ARGS")
+            print("用法：python main.py generate <平台名或account_id> \"<提示词>\"")
+            return 1
+        if not cmd_generate(args[1], args[2]):
+            return 1
+
+    elif cmd == "list":
+        platform_filter, limit, err = _parse_list_args(args[1:])
+        if err:
+            return err
+        cmd_key_list(platform_filter, limit=limit)
+
+    elif cmd == "add":
+        platform_arg, api_key_arg, model_arg, label_arg, err = _parse_add_args(args[1:])
+        if err:
+            return err
+        cmd_key_add(platform_arg, api_key_arg, model=model_arg, label=label_arg)
+
+    elif cmd == "del":
+        if len(args) < 2:
+            print("ERROR:CLI_DEL_MISSING_ARGS 用法：python main.py del <key_id>")
+            return 1
+        cmd_key_del(args[1])
+
+    elif cmd == "key":
+        if len(args) < 2:
+            print("ERROR:CLI_KEY_MISSING_SUBCOMMAND 请指定子命令：add / list / del")
+            return 1
+        sub = args[1]
+
+        if sub == "add":
+            platform_arg, api_key_arg, model_arg, label_arg, err = _parse_add_args(args[2:])
+            if err:
+                return err
+            cmd_key_add(platform_arg, api_key_arg, model=model_arg, label=label_arg)
+
+        elif sub == "list":
+            platform_filter, limit, err = _parse_list_args(args[2:])
+            if err:
+                return err
+            cmd_key_list(platform_filter, limit=limit)
+
+        elif sub == "del":
+            if len(args) < 3:
+                print("ERROR:CLI_KEY_DEL_MISSING_ARGS 用法：python main.py key del <key_id>")
+                return 1
+            cmd_key_del(args[2])
+
+        else:
+            print(f"ERROR:CLI_UNKNOWN_KEY_SUB 未知 key 子命令「{sub}」，支持：add / list / del")
+            return 1
+
+    else:
+        print(f"ERROR:CLI_UNKNOWN_COMMAND 未知命令「{cmd}」。")
+        _print_usage()
+        return 1
+
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

llm-manager/scripts/providers.py

@@ -0,0 +1,199 @@
+"""
+平台配置中心：7 个 LLM 平台的静态配置、别名解析、以及通过 account-manager 暴露的 CLI 查询已登录网页账号。
+"""
+import json
+import os
+import subprocess
+import sys
+
+# ---------------------------------------------------------------------------
+# 平台静态配置
+# ---------------------------------------------------------------------------
+LLM_PROVIDERS = {
+    "doubao": {
+        "label": "豆包",
+        "aliases": ["豆包"],
+        "web_url": "https://www.doubao.com/chat/",
+        "api_base": "https://ark.volces.com/api/v3",
+        "api_model": None,          # 豆包需要用户在火山引擎控制台建推理接入点，model=ep-xxx
+        "has_api": True,
+        "api_note": "需先在火山引擎控制台创建推理接入点，--model 传 endpoint_id（格式 ep-xxx）",
+    },
+    "deepseek": {
+        "label": "DeepSeek",
+        "aliases": ["深度求索"],
+        "web_url": "https://chat.deepseek.com",
+        "api_base": "https://api.deepseek.com/v1",
+        "api_model": "deepseek-chat",
+        "has_api": True,
+        "api_note": "模型可选：deepseek-chat / deepseek-reasoner",
+    },
+    "qianwen": {
+        "label": "通义千问",
+        "aliases": ["通义", "千问", "qwen", "tongyi"],
+        "web_url": "https://tongyi.aliyun.com/qianwen/",
+        "api_base": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+        "api_model": "qwen-plus",
+        "has_api": True,
+        "api_note": "模型可选：qwen-turbo / qwen-plus / qwen-max",
+    },
+    "kimi": {
+        "label": "Kimi",
+        "aliases": ["月之暗面", "moonshot"],
+        "web_url": "https://kimi.moonshot.cn",
+        "api_base": "https://api.moonshot.cn/v1",
+        "api_model": "moonshot-v1-8k",
+        "has_api": True,
+        "api_note": "模型可选：moonshot-v1-8k / moonshot-v1-32k / moonshot-v1-128k",
+    },
+    "yiyan": {
+        "label": "文心一言",
+        "aliases": ["文心", "一言", "ernie", "wenxin"],
+        "web_url": "https://yiyan.baidu.com",
+        "api_base": "https://qianfan.baidubce.com/v2",
+        "api_model": "ernie-4.0-8k",
+        "has_api": True,
+        "api_note": "模型可选：ernie-4.0-8k / ernie-3.5-8k",
+    },
+    "yuanbao": {
+        "label": "腾讯元宝",
+        "aliases": ["元宝"],
+        "web_url": "https://yuanbao.tencent.com/chat",
+        "api_base": None,
+        "api_model": None,
+        "has_api": False,
+        "api_note": "暂无公开 API，仅支持网页模式",
+    },
+    "minimax": {
+        "label": "MiniMax",
+        "aliases": ["minimax", "MiniMax", "海螺", "海螺AI"],
+        "web_url": "https://chat.minimax.io/",
+        "api_base": "https://api.minimax.chat/v1",
+        "api_model": "MiniMax-Text-01",
+        "has_api": True,
+        "api_note": "模型可按 MiniMax 控制台可用模型调整，建议通过 --model 显式指定。",
+    },
+}
+
+# 构建别名查找表（含中文、英文键、aliases）
+_ALIAS_TO_KEY: dict = {}
+for _k, _spec in LLM_PROVIDERS.items():
+    _ALIAS_TO_KEY[_k] = _k
+    _ALIAS_TO_KEY[_k.lower()] = _k
+    _ALIAS_TO_KEY[_spec["label"]] = _k
+    for _a in (_spec.get("aliases") or []):
+        _ALIAS_TO_KEY[_a] = _k
+        _ALIAS_TO_KEY[_a.lower()] = _k
+
+
+def resolve_provider_key(name: str):
+    """将用户输入的平台名称/别名解析为内部 slug；无法识别返回 None。"""
+    if not name:
+        return None
+    s = str(name).strip()
+    return _ALIAS_TO_KEY.get(s) or _ALIAS_TO_KEY.get(s.lower())
+
+
+def provider_list_cn() -> str:
+    return "、".join(s["label"] for s in LLM_PROVIDERS.values())
+
+
+# ---------------------------------------------------------------------------
+# 路径帮助（与 account-manager/scripts/main.py 完全一致：仅 JIANGCHANG_*）
+# ---------------------------------------------------------------------------
+
+_OPENCLAW_DIR = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+
+def get_data_root() -> str:
+    env = (os.getenv("JIANGCHANG_DATA_ROOT") or "").strip()
+    if env:
+        return env
+    if sys.platform == "win32":
+        return r"D:\jiangchang-data"
+    return os.path.join(os.path.expanduser("~"), ".jiangchang-data")
+
+
+def get_user_id() -> str:
+    uid = (os.getenv("JIANGCHANG_USER_ID") or "").strip()
+    return uid or "_anon"
+
+
+# ---------------------------------------------------------------------------
+# 跨技能：仅通过 account-manager 提供的 CLI 读取账号（不直接打开其数据库文件）
+# ---------------------------------------------------------------------------
+
+def _account_manager_script_path() -> str:
+    return os.path.join(_OPENCLAW_DIR, "account-manager", "scripts", "main.py")
+
+
+def find_logged_in_account(provider_key: str) -> dict | None:
+    """
+    调用 account-manager：pick-logged-in <platform_key>，取该平台已登录（login_status=1）的优先账号。
+    成功返回与 account.py get 一致的 dict；失败或不可用时返回 None。
+    """
+    script = _account_manager_script_path()
+    if not os.path.isfile(script):
+        return None
+    try:
+        proc = subprocess.run(
+            [sys.executable, script, "pick-logged-in", provider_key],
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+        )
+    except OSError:
+        return None
+    raw = (proc.stdout or "").strip()
+    if not raw or raw.startswith("ERROR:"):
+        return None
+    try:
+        data = json.loads(raw.splitlines()[0])
+    except (json.JSONDecodeError, IndexError):
+        return None
+    if not isinstance(data, dict) or data.get("id") is None:
+        return None
+    plat = data.get("platform") or provider_key
+    if not (data.get("url") or "").strip():
+        data["url"] = LLM_PROVIDERS.get(provider_key, {}).get("web_url", "")
+    data["platform"] = plat
+    return data
+
+
+# ---------------------------------------------------------------------------
+# Chrome/Edge 检测（与 account-manager 逻辑保持一致）
+# ---------------------------------------------------------------------------
+
+def _win_find_exe(candidates):
+    for p in candidates:
+        if p and os.path.isfile(p):
+            return p
+    return None
+
+
+def resolve_chromium_channel() -> str | None:
+    """返回 'chrome' | 'msedge' | None。"""
+    if sys.platform != "win32":
+        return "chrome"
+    pf = os.environ.get("ProgramFiles", r"C:\Program Files")
+    pfx86 = os.environ.get("ProgramFiles(x86)", r"C:\Program Files (x86)")
+    local = os.environ.get("LocalAppData", "")
+
+    chrome = _win_find_exe([
+        os.path.join(pf, "Google", "Chrome", "Application", "chrome.exe"),
+        os.path.join(pfx86, "Google", "Chrome", "Application", "chrome.exe"),
+        os.path.join(local, "Google", "Chrome", "Application", "chrome.exe") if local else "",
+    ])
+    if chrome:
+        return "chrome"
+
+    edge = _win_find_exe([
+        os.path.join(pfx86, "Microsoft", "Edge", "Application", "msedge.exe"),
+        os.path.join(pf, "Microsoft", "Edge", "Application", "msedge.exe"),
+        os.path.join(local, "Microsoft", "Edge", "Application", "msedge.exe") if local else "",
+    ])
+    if edge:
+        return "msedge"
+
+    return None

logistics-tracker/.github/workflows/release_skill.yaml

@@ -0,0 +1,11 @@
+name: 技能自动化发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

logistics-tracker/SKILL.md

@@ -0,0 +1,53 @@
+---
+name: 物流轨迹查询
+description: 物流轨迹查询。当用户发送物流单号、询问包裹状态、问"货到哪了"、"帮我查单号"等时使用。调用17Track API查询最新轨迹并用中文回复。支持全球200+承运商。
+version: 1.0.0
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: logistics-tracker
+    emoji: "📦"
+    category: "通用"
+    requires:
+      env:
+        - TRACK17_API_KEY
+      bins:
+        - python3
+allowed-tools:
+  - bash
+---
+
+# 物流轨迹查询
+
+## 使用时机
+
+当用户发送以下内容时触发本Skill：
+- 包含物流单号（字母+数字组合，如 `RR123456789CN`、`UPS1234567890`）
+- 询问"我的货到哪了"、"帮我查单号"、"物流查询"、"查一下快递"等
+
+## 执行步骤
+
+1. 从用户消息中提取物流单号
+2. 执行查询脚本：
+```bash
+python3 {baseDir}/scripts/main.py <单号>
+```
+
+3. 将脚本返回结果直接回复给用户
+4. 如果用户没有提供单号，回复："请发送您要查询的物流单号，我来帮您查询最新状态。"
+
+## 输出格式
+
+脚本会返回格式化好的中文结果，直接发送给用户即可，无需二次处理。
+
+## 错误处理
+
+- 查询失败：告知用户"该单号暂时查询不到，可能未入网，请稍后再试"
+- 网络超时：告知用户"网络连接超时，请稍后重试"
+- 无效单号：告知用户"单号格式不正确，请检查后重新发送"
+
+## 注意事项
+
+- 每次只处理一个单号
+- API Key 通过环境变量 `TRACK17_API_KEY` 读取，不要硬编码在脚本里
+- 免费套餐每月100次查询，正式使用需升级

logistics-tracker/release.ps1

@@ -0,0 +1,220 @@
+<#
+.SYNOPSIS
+  One-command release script for skill repos.
+
+.DESCRIPTION
+  - Optional auto-commit
+  - Push current branch
+  - Auto-increment semantic tag (vX.Y.Z)
+  - Create & push tag
+  - Fail fast on unsafe states
+
+.EXAMPLES
+  # Safe mode (recommended): requires clean working tree
+  .\release.ps1
+
+  # Auto commit tracked/untracked changes then release
+  .\release.ps1 -AutoCommit -CommitMessage "chore: update skill config"
+
+  # Dry run (show what would happen)
+  .\release.ps1 -DryRun
+
+  # Custom tag prefix
+  .\release.ps1 -Prefix "v" -Message "正式发布"
+
+.NOTES
+  Requires: git, PowerShell 5+
+#>
+
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+if (Test-Path $sharedScript) {
+  & $sharedScript @PSBoundParameters
+  exit $LASTEXITCODE
+}
+
+function Invoke-Git {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  Write-Host ">> git $Args" -ForegroundColor DarkGray
+  & cmd /c "git $Args"
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+}
+
+function Get-GitOutput {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  $output = & cmd /c "git $Args" 2>$null
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+  return @($output)
+}
+
+function Test-Repo {
+  & git rev-parse --is-inside-work-tree *> $null
+  return ($LASTEXITCODE -eq 0)
+}
+
+function Get-CurrentBranch {
+  $b = (Get-GitOutput "branch --show-current" | Select-Object -First 1).Trim()
+  return $b
+}
+
+function Get-StatusPorcelain {
+  $lines = @(Get-GitOutput "status --porcelain")
+  return $lines
+}
+
+function Parse-SemVerTag {
+  param(
+    [string]$Tag,
+    [string]$TagPrefix
+  )
+  $escaped = [regex]::Escape($TagPrefix)
+  $m = [regex]::Match($Tag, "^${escaped}(\d+)\.(\d+)\.(\d+)$")
+  if (-not $m.Success) { return $null }
+
+  return [pscustomobject]@{
+    Raw   = $Tag
+    Major = [int]$m.Groups[1].Value
+    Minor = [int]$m.Groups[2].Value
+    Patch = [int]$m.Groups[3].Value
+  }
+}
+
+function Get-NextTag {
+  param([string]$TagPrefix)
+
+  $tags = Get-GitOutput "tag --list"
+  $parsed = @()
+
+  foreach ($t in $tags) {
+    $t = $t.Trim()
+    if (-not $t) { continue }
+    $obj = Parse-SemVerTag -Tag $t -TagPrefix $TagPrefix
+    if ($null -ne $obj) { $parsed += $obj }
+  }
+
+  if ($parsed.Count -eq 0) {
+    return "${TagPrefix}1.0.1"
+  }
+
+  $latest = $parsed | Sort-Object Major, Minor, Patch | Select-Object -Last 1
+  return "$TagPrefix$($latest.Major).$($latest.Minor).$([int]$latest.Patch + 1)"
+}
+
+function Ensure-CleanOrAutoCommit {
+  param(
+    [switch]$DoAutoCommit,
+    [switch]$NeedClean,
+    [switch]$IsDryRun,
+    [string]$Msg
+  )
+
+  $status = @(Get-StatusPorcelain)
+  if ($status.Length -eq 0) { return }
+
+  if ($NeedClean) {
+    Write-Host "Working tree is not clean and -RequireClean is enabled." -ForegroundColor Yellow
+    & git status --short
+    throw "Abort: dirty working tree."
+  }
+
+  # 默认一键发布：有改动就自动提交；也可用 -AutoCommit 显式开启
+  $commitMsg = $Msg
+  if ([string]::IsNullOrWhiteSpace($commitMsg)) {
+    $commitMsg = "chore: auto release commit ($(Get-Date -Format 'yyyy-MM-dd HH:mm:ss'))"
+  }
+
+  if (-not $DoAutoCommit) {
+    Write-Host "Detected uncommitted changes, auto-committing before release..." -ForegroundColor Yellow
+  }
+
+  if ($IsDryRun) {
+    Write-Host "[DryRun] Would run: git add -A" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git commit -m `"$commitMsg`"" -ForegroundColor Yellow
+    return
+  }
+
+  Invoke-Git "add -A"
+  Invoke-Git "commit -m `"$commitMsg`""
+}
+
+try {
+  Write-Host "=== Release Script Start ===" -ForegroundColor Cyan
+
+  if (-not (Test-Repo)) {
+    throw "Current directory is not a git repository."
+  }
+
+  $branch = Get-CurrentBranch
+  if ([string]::IsNullOrWhiteSpace($branch)) {
+    throw "Unable to determine current branch."
+  }
+
+  if ($branch -notin @("main", "master")) {
+    throw "Current branch is '$branch'. Release is only allowed from main/master."
+  }
+
+  Invoke-Git "fetch --tags --prune origin"
+
+  Ensure-CleanOrAutoCommit -DoAutoCommit:$AutoCommit -NeedClean:$RequireClean -IsDryRun:$DryRun -Msg $CommitMessage
+
+  $upstream = (& git rev-parse --abbrev-ref --symbolic-full-name "@{u}" 2>$null)
+  $hasUpstream = ($LASTEXITCODE -eq 0)
+
+  if ($DryRun) {
+    if ($hasUpstream) {
+      Write-Host "[DryRun] Would run: git push" -ForegroundColor Yellow
+    } else {
+      Write-Host "[DryRun] Would run: git push -u origin $branch" -ForegroundColor Yellow
+    }
+  } else {
+    if ($hasUpstream) {
+      Invoke-Git "push"
+    } else {
+      Invoke-Git "push -u origin $branch"
+    }
+  }
+
+  $nextTag = Get-NextTag -TagPrefix $Prefix
+  Write-Host "Next tag: $nextTag" -ForegroundColor Green
+
+  $existing = @(Get-GitOutput "tag --list `"$nextTag`"")
+  if ($existing.Length -gt 0) {
+    throw "Tag already exists: $nextTag"
+  }
+
+  if ($DryRun) {
+    Write-Host "[DryRun] Would run: git tag -a $nextTag -m `"$Message`"" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git push origin $nextTag" -ForegroundColor Yellow
+    Write-Host "=== DryRun Complete ===" -ForegroundColor Cyan
+    exit 0
+  }
+
+  Invoke-Git "tag -a $nextTag -m `"$Message`""
+  Invoke-Git "push origin $nextTag"
+
+  Write-Host "Release success: $nextTag" -ForegroundColor Green
+  Write-Host "=== Release Script Done ===" -ForegroundColor Cyan
+  exit 0
+}
+catch {
+  Write-Host "Release failed: $($_.Exception.Message)" -ForegroundColor Red
+  exit 1
+}

logistics-tracker/scripts/main.py

@@ -0,0 +1,142 @@
+import requests
+import sys
+import os
+import subprocess
+
+
+def get_api_key(key_name):
+    """从api-key-vault读取Key"""
+    vault_path = os.path.join(
+        os.path.dirname(__file__),
+        "..", "..",          # 从scripts/退到logistics-tracker/再退到OpenClaw/
+        "api-key-vault", "scripts", "vault.py"
+    )
+    vault_path = os.path.normpath(vault_path)
+
+    result = subprocess.run(
+        ["python", vault_path, "get", key_name],
+        capture_output=True, text=True
+    )
+    key = result.stdout.strip()
+
+    if not key or key == "ERROR:KEY_NOT_FOUND":
+        return None
+    return key
+
+
+def check_entitlement(skill_slug):
+    auth_base = (os.getenv("JIANGCHANG_AUTH_BASE_URL") or "").strip().rstrip("/")
+    if not auth_base:
+        return True, ""
+
+    user_id = (os.getenv("JIANGCHANG_USER_ID") or "").strip()
+    if not user_id:
+        return False, "鉴权失败：缺少用户身份（JIANGCHANG_USER_ID）"
+
+    auth_api_key = (os.getenv("JIANGCHANG_AUTH_API_KEY") or "").strip()
+    timeout = int((os.getenv("JIANGCHANG_AUTH_TIMEOUT_SECONDS") or "5").strip())
+    headers = {"Content-Type": "application/json"}
+    if auth_api_key:
+        headers["Authorization"] = f"Bearer {auth_api_key}"
+
+    payload = {
+        "user_id": user_id,
+        "skill_slug": skill_slug,
+        "trace_id": (os.getenv("JIANGCHANG_TRACE_ID") or "").strip(),
+        "context": {"entry": "main.py"},
+    }
+    try:
+        res = requests.post(
+            f"{auth_base}/api/entitlements/check",
+            json=payload,
+            headers=headers,
+            timeout=timeout,
+        )
+    except requests.RequestException as exc:
+        return False, f"鉴权请求失败：{exc}"
+
+    if res.status_code != 200:
+        return False, f"鉴权服务异常：HTTP {res.status_code}"
+    try:
+        body = res.json()
+    except ValueError:
+        return False, "鉴权服务异常：返回非 JSON"
+
+    code = body.get("code")
+    data = body.get("data") or {}
+    if code != 200:
+        return False, str(body.get("msg") or "鉴权失败")
+    if not data.get("allow", False):
+        return False, str(data.get("reason") or "未购买或已过期")
+    return True, ""
+
+
+def query_tracking(tracking_number):
+    api_key = get_api_key("17track")
+    if not api_key:
+        return "错误：未找到17track的API Key，请先运行：\npython api-key-vault/scripts/vault.py set 17track 你的Key"
+
+    url = "https://api.17track.net/track/v2.2/gettrackinfo"
+    headers = {
+        "Content-Type": "application/json",
+        "17token": api_key
+    }
+    body = {
+        "data": [{"number": tracking_number}]
+    }
+
+    try:
+        response = requests.post(url, json=body, headers=headers, timeout=10)
+        result = response.json()
+
+        accepted = result.get("data", {}).get("accepted", [])
+        if not accepted:
+            return f"抱歉，单号 {tracking_number} 查询不到信息，可能还未入网，请稍后再试。"
+
+        track_info = accepted[0].get("track", {})
+        providers = track_info.get("tracking", {}).get("providers", [])
+
+        if not providers or not providers[0].get("events"):
+            return f"📦 单号：{tracking_number}\n该单号已入网，暂无轨迹更新，请稍后再查。"
+
+        events = providers[0]["events"]
+        latest = events[0]
+
+        recent_lines = []
+        for e in events[:3]:
+            time = e.get("time_iso", "")
+            location = e.get("location", "") or "未知"
+            desc = e.get("description", "")
+            recent_lines.append(f"  · {time}  {location}  {desc}")
+
+        recent_text = "\n".join(recent_lines)
+
+        return (
+            f"📦 单号：{tracking_number}\n"
+            f"📍 最新状态：{latest.get('description', '未知')}\n"
+            f"🕐 更新时间：{latest.get('time_iso', '未知')}\n"
+            f"🗺  最新位置：{latest.get('location', '未知')}\n\n"
+            f"最近轨迹：\n{recent_text}"
+        )
+
+    except requests.exceptions.Timeout:
+        return "查询超时，请稍后重试。"
+    except Exception as e:
+        return f"查询失败：{str(e)}"
+
+
+def main(argv=None) -> int:
+    args = argv if argv is not None else sys.argv[1:]
+    if len(args) < 1:
+        print("用法：python main.py <单号>")
+        return 1
+    ok, reason = check_entitlement("logistics-tracker")
+    if not ok:
+        print(f"❌ {reason}")
+        return 1
+    print(query_tracking(args[0]))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

skill-template/README.md

@@ -16,12 +16,14 @@
 4. **扩展子命令**：在 `scripts/skill_main.py` 的 `dispatch` 中增加分支；业务逻辑放在同目录其它模块或子包中，保持入口轻薄。
 5. **编写/调整 `SKILL.md`**：只改「何时触发、如何调用、参数含义」，不要写实现细节；实现细节放在 `docs/` 或代码注释里。
 6. **发布**：若使用 GitHub Actions，编辑 `.github/workflows/release_skill.yaml`，把 `uses:` 指向**你们组织**的复用工作流；若不用 CI，可删除该目录。
+7. **一键打标签推送（与匠厂 monorepo 对齐）**：在技能仓库根目录执行 `.\release.ps1`（需与 `jiangchang-platform-kit` 位于同一父目录，以便调用 `..\jiangchang-platform-kit\tools\release.ps1`）。支持 `-DryRun`、`-AutoCommit`、`-CommitMessage` 等参数，与 `account-manager` / `sohu-publisher` 一致。
 
 ## 目录一览
 
 | 路径 | 作用 |
 |------|------|
 | `SKILL.md` | 技能清单（YAML 头 + Markdown 正文），供宿主与协作者阅读 |
+| `release.ps1` | 转调平台套件的发布脚本（提交/推送/语义化 tag）；依赖并列的 `jiangchang-platform-kit` |
 | `scripts/skill_main.py` | 推荐唯一 CLI 入口；含 `health` / `version` 示例 |
 | `docs/RUNTIME.md` | 环境与目录契约（多宿主通用） |
 | `docs/SKILL_TYPES.md` | 常见技能形态与自检清单 |

skill-template/docs/LOGGING.md

@@ -0,0 +1,59 @@
+# OpenClaw 技能日志约定（各技能统一）
+
+本文档为**规范**：新技能与 `account-manager` 等现有技能应遵循同一套路径、格式与环境变量，便于排查与运维。
+
+## 目录与文件
+
+- 日志目录与技能数据目录同级：
+  - `{JIANGCHANG_DATA_ROOT}/{JIANGCHANG_USER_ID}/{skill_slug}/logs/`
+- 主日志文件（示例）：
+  - `{...}/account-manager/logs/account-manager.log`
+- `skill_slug` 与 `SKILL.md` 中 `metadata.openclaw.slug` 一致。
+
+## 轮转与编码
+
+- 使用 Python `logging.handlers.TimedRotatingFileHandler`，`when=midnight`，保留约 30 个历史文件（可按技能调整 `backupCount`）。
+- 文件编码 **UTF-8**。
+- 首次写日志前创建 `logs/` 目录（`exist_ok=True`）。
+
+## 日志格式（行文本）
+
+推荐单行格式（与 `account-manager` 一致）：
+
+```text
+%(asctime)s | %(levelname)-8s | %(name)s | %(message)s
+```
+
+`datefmt` 建议：`%Y-%m-%dT%H:%M:%S`（本地时间）。
+
+## Logger 命名
+
+- 使用分层命名，便于按前缀过滤，例如：`openclaw.skill.account_manager`。
+- 子进程若需写同一文件，可使用子 logger：`openclaw.skill.account_manager.login_child`，仍追加到**同一日志文件**（`logging.FileHandler(..., encoding='utf-8')` 追加模式）。
+
+## 环境变量（全局）
+
+| 变量 | 说明 |
+|------|------|
+| `JIANGCHANG_LOG_LEVEL` | 默认 `INFO`；调试轮询等可设 `DEBUG`。 |
+| `JIANGCHANG_LOG_TO_STDERR` | 设为 `1`/`true` 时，将 `WARNING` 及以上同时输出到 `stderr`（不替代文件）。 |
+| `JIANGCHANG_{SKILL}_LOG_FILE` | 可选，覆盖该技能主日志**绝对路径**（需自行保证目录可写）。`account-manager` 使用 `JIANGCHANG_ACCOUNT_MANAGER_LOG_FILE`。 |
+
+## 级别与内容建议
+
+- **INFO**：子命令入口、成功结束（含关键业务 id，避免仅打「成功」）。
+- **WARNING**：可恢复问题、校验失败、登录未检测到。
+- **ERROR**：依赖缺失、不可继续的异常。
+- **DEBUG**：高频轮询状态（URL 列表、规则命中说明）；默认关闭。
+- 日志中可能含 **手机号、路径等敏感信息**，注意文件权限与留存策略；必要时仅记录后四位或 hash（按合规要求）。
+
+## 与标准输出的关系
+
+- 面向用户的 `print`（成功提示、`ERROR:` 前缀）可保留；**诊断细节以日志为准**。
+- 登录失败等场景可在终端提示日志文件路径，便于用户打开 `*.log`。
+
+## 参考实现
+
+- `account-manager/scripts/main.py`：`get_skill_logs_dir`、`get_skill_logger`、`get_skill_log_file_path`、login 子进程内 `FileHandler` 追加同一文件。
+
+新技能可复制上述函数结构，替换 `SKILL_SLUG` 与 `LOG_LOGGER_NAME` 即可。

skill-template/release.ps1

@@ -0,0 +1,23 @@
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+
+if (-not (Test-Path $sharedScript)) {
+  throw "Shared release script not found: $sharedScript"
+}
+
+& $sharedScript @PSBoundParameters
+exit $LASTEXITCODE

sohu-publisher/.github/workflows/release_skill.yaml

@@ -0,0 +1,11 @@
+name: 技能自动化发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

sohu-publisher/SKILL.md

@@ -0,0 +1,77 @@
+---
+name: 搜狐号自动发布
+description: 搜狐号自动发布大总管。系统会自动帮你查找文章ID和账号ID，严禁向用户询问代码层的参数！
+version: 1.1.0
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: sohu-publisher
+    emoji: "📰"
+    category: "通用"
+    dependencies:
+      required:
+        - content-manager
+        - account-manager
+      auto_install: true
+allowed-tools:
+  - bash
+---
+
+# 搜狐号全自动发布指引
+
+## 🚨 AI 必备纪律 (严禁直接询问用户参数 ❗)
+
+你是一个面向小白用户的顶级全自动运营总监。用户**不懂代码**，他们不知道什么是 `article_id=3` ，也不懂什么是 `account_id=sohu_account1`。
+用户只会用自己的视角跟你下达命令（比如：“帮我把最新的一篇发到搜狐号”、“用 13800000000 那个账号发一下文章”）。
+
+当你收到这类含有发布意图的指令时，**绝对不能打断用户，要求他们提供 account_id 和 article_id**。你需要按照以下【静默 3 步走】机制，替用户去把需要的值找出来！
+
+### 🎯 步骤 1：自动感知识别【文章数字 ID】
+如果你不知道确切的文章ID数字是多少（比如用户只说了：发下最新的，或者发刚写好的）：
+1. 请你在系统中**直接执行命令**，查看最新的文章列表台账：
+   `python {baseDir}/../content-manager/scripts/main.py list`
+2. 阅读返回的表格。通过列表最顶部的记录，或者是标题匹配您的上下文，**你自己**把那篇目标文章的最左侧**数字 ID**挖出来。千万别去问用户。
+
+### 🎯 步骤 2：自动感知识别【账号 ID】
+用户通常用**手机号**区分账号。如果你不知道该用哪个发：
+1. 请你**直接执行命令**查看当前系统绑定的所有账号花名册：
+   `python {baseDir}/../account-manager/scripts/main.py list`
+2. 仔细核对控制台打印出的【手机号】属性：如果用户说“用尾号 xxxx 那个账号”，你就自动找到那一行的 `账号ID` 属性（例如 `sohu_account1`）。
+3. **特例**：如果你发现其实系统里只绑定了 1 个搜狐账号，你连查都不用查，直接默认提取 `sohu_account1` 充当参数。绝对不要拿这种事麻烦用户！
+
+### 🎯 步骤 3：直接一键开火发布
+当你自己在后台摸清楚了 `account_id`（字符串）和 `article_id`（数字） 后，请热情地对用户说：“好的老大，这就开始帮您操作发布...”，然后立刻替用户在终端执以下命令发起物理发布引擎：
+
+```bash
+python {baseDir}/scripts/main.py <account_id> <article_id>
+```
+
+**示例**：
+`python {baseDir}/scripts/main.py sohu_account1 3`
+
+## 常用 CLI（给小白也能看懂）
+
+```bash
+# 发布一篇（推荐显式写 publish）
+python {baseDir}/scripts/main.py publish <account_id> <article_id>
+
+# 兼容旧写法（等价于 publish）
+python {baseDir}/scripts/main.py <account_id> <article_id>
+
+# 查看发布记录（默认最新 10 条，按创建时间倒序）
+python {baseDir}/scripts/main.py logs
+python {baseDir}/scripts/main.py logs --limit 20
+python {baseDir}/scripts/main.py logs --status failed
+python {baseDir}/scripts/main.py logs --account-id sohu_account1
+
+# 查看某一条发布记录（JSON）
+python {baseDir}/scripts/main.py log-get <log_id>
+
+# 健康/版本
+python {baseDir}/scripts/main.py health
+python {baseDir}/scripts/main.py version
+```
+
+---
+> 💡 若你执行发布脚本后，控制台向你抛出了 `ERROR:REQUIRE_LOGIN` 警告和一系列提示。这说明你成功启动了代理人流程，但搜狐账号已经掉线了！这时脚本实际上已经替您调起了浏览器的登录弹窗。
+> 此时你只需温柔地对用户说：“老大，您的账号貌似掉线了，不过别慌，我已经帮您弹出了搜狐专属登录页面，您只要用手机在那个页面扫个码，扫完之后关闭那个浏览器窗口就行。等您弄好了告诉我，我再帮您从头发布一次！”

sohu-publisher/release.ps1

@@ -0,0 +1,220 @@
+<#
+.SYNOPSIS
+  One-command release script for skill repos.
+
+.DESCRIPTION
+  - Optional auto-commit
+  - Push current branch
+  - Auto-increment semantic tag (vX.Y.Z)
+  - Create & push tag
+  - Fail fast on unsafe states
+
+.EXAMPLES
+  # Safe mode (recommended): requires clean working tree
+  .\release.ps1
+
+  # Auto commit tracked/untracked changes then release
+  .\release.ps1 -AutoCommit -CommitMessage "chore: update skill config"
+
+  # Dry run (show what would happen)
+  .\release.ps1 -DryRun
+
+  # Custom tag prefix
+  .\release.ps1 -Prefix "v" -Message "正式发布"
+
+.NOTES
+  Requires: git, PowerShell 5+
+#>
+
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+if (Test-Path $sharedScript) {
+  & $sharedScript @PSBoundParameters
+  exit $LASTEXITCODE
+}
+
+function Invoke-Git {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  Write-Host ">> git $Args" -ForegroundColor DarkGray
+  & cmd /c "git $Args"
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+}
+
+function Get-GitOutput {
+  param([Parameter(Mandatory = $true)][string]$Args)
+  $output = & cmd /c "git $Args" 2>$null
+  if ($LASTEXITCODE -ne 0) {
+    throw "git command failed: git $Args"
+  }
+  return @($output)
+}
+
+function Test-Repo {
+  & git rev-parse --is-inside-work-tree *> $null
+  return ($LASTEXITCODE -eq 0)
+}
+
+function Get-CurrentBranch {
+  $b = (Get-GitOutput "branch --show-current" | Select-Object -First 1).Trim()
+  return $b
+}
+
+function Get-StatusPorcelain {
+  $lines = @(Get-GitOutput "status --porcelain")
+  return $lines
+}
+
+function Parse-SemVerTag {
+  param(
+    [string]$Tag,
+    [string]$TagPrefix
+  )
+  $escaped = [regex]::Escape($TagPrefix)
+  $m = [regex]::Match($Tag, "^${escaped}(\d+)\.(\d+)\.(\d+)$")
+  if (-not $m.Success) { return $null }
+
+  return [pscustomobject]@{
+    Raw   = $Tag
+    Major = [int]$m.Groups[1].Value
+    Minor = [int]$m.Groups[2].Value
+    Patch = [int]$m.Groups[3].Value
+  }
+}
+
+function Get-NextTag {
+  param([string]$TagPrefix)
+
+  $tags = Get-GitOutput "tag --list"
+  $parsed = @()
+
+  foreach ($t in $tags) {
+    $t = $t.Trim()
+    if (-not $t) { continue }
+    $obj = Parse-SemVerTag -Tag $t -TagPrefix $TagPrefix
+    if ($null -ne $obj) { $parsed += $obj }
+  }
+
+  if ($parsed.Count -eq 0) {
+    return "${TagPrefix}1.0.1"
+  }
+
+  $latest = $parsed | Sort-Object Major, Minor, Patch | Select-Object -Last 1
+  return "$TagPrefix$($latest.Major).$($latest.Minor).$([int]$latest.Patch + 1)"
+}
+
+function Ensure-CleanOrAutoCommit {
+  param(
+    [switch]$DoAutoCommit,
+    [switch]$NeedClean,
+    [switch]$IsDryRun,
+    [string]$Msg
+  )
+
+  $status = @(Get-StatusPorcelain)
+  if ($status.Length -eq 0) { return }
+
+  if ($NeedClean) {
+    Write-Host "Working tree is not clean and -RequireClean is enabled." -ForegroundColor Yellow
+    & git status --short
+    throw "Abort: dirty working tree."
+  }
+
+  # 默认一键发布：有改动就自动提交；也可用 -AutoCommit 显式开启
+  $commitMsg = $Msg
+  if ([string]::IsNullOrWhiteSpace($commitMsg)) {
+    $commitMsg = "chore: auto release commit ($(Get-Date -Format 'yyyy-MM-dd HH:mm:ss'))"
+  }
+
+  if (-not $DoAutoCommit) {
+    Write-Host "Detected uncommitted changes, auto-committing before release..." -ForegroundColor Yellow
+  }
+
+  if ($IsDryRun) {
+    Write-Host "[DryRun] Would run: git add -A" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git commit -m `"$commitMsg`"" -ForegroundColor Yellow
+    return
+  }
+
+  Invoke-Git "add -A"
+  Invoke-Git "commit -m `"$commitMsg`""
+}
+
+try {
+  Write-Host "=== Release Script Start ===" -ForegroundColor Cyan
+
+  if (-not (Test-Repo)) {
+    throw "Current directory is not a git repository."
+  }
+
+  $branch = Get-CurrentBranch
+  if ([string]::IsNullOrWhiteSpace($branch)) {
+    throw "Unable to determine current branch."
+  }
+
+  if ($branch -notin @("main", "master")) {
+    throw "Current branch is '$branch'. Release is only allowed from main/master."
+  }
+
+  Invoke-Git "fetch --tags --prune origin"
+
+  Ensure-CleanOrAutoCommit -DoAutoCommit:$AutoCommit -NeedClean:$RequireClean -IsDryRun:$DryRun -Msg $CommitMessage
+
+  $upstream = (& git rev-parse --abbrev-ref --symbolic-full-name "@{u}" 2>$null)
+  $hasUpstream = ($LASTEXITCODE -eq 0)
+
+  if ($DryRun) {
+    if ($hasUpstream) {
+      Write-Host "[DryRun] Would run: git push" -ForegroundColor Yellow
+    } else {
+      Write-Host "[DryRun] Would run: git push -u origin $branch" -ForegroundColor Yellow
+    }
+  } else {
+    if ($hasUpstream) {
+      Invoke-Git "push"
+    } else {
+      Invoke-Git "push -u origin $branch"
+    }
+  }
+
+  $nextTag = Get-NextTag -TagPrefix $Prefix
+  Write-Host "Next tag: $nextTag" -ForegroundColor Green
+
+  $existing = @(Get-GitOutput "tag --list `"$nextTag`"")
+  if ($existing.Length -gt 0) {
+    throw "Tag already exists: $nextTag"
+  }
+
+  if ($DryRun) {
+    Write-Host "[DryRun] Would run: git tag -a $nextTag -m `"$Message`"" -ForegroundColor Yellow
+    Write-Host "[DryRun] Would run: git push origin $nextTag" -ForegroundColor Yellow
+    Write-Host "=== DryRun Complete ===" -ForegroundColor Cyan
+    exit 0
+  }
+
+  Invoke-Git "tag -a $nextTag -m `"$Message`""
+  Invoke-Git "push origin $nextTag"
+
+  Write-Host "Release success: $nextTag" -ForegroundColor Green
+  Write-Host "=== Release Script Done ===" -ForegroundColor Cyan
+  exit 0
+}
+catch {
+  Write-Host "Release failed: $($_.Exception.Message)" -ForegroundColor Red
+  exit 1
+}

sohu-publisher/scripts/main.py

@@ -0,0 +1,563 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+sohu-publisher：搜狐号自动发布。
+
+子命令：
+  publish <account_id> <article_id>       发布文章
+  logs [--limit N] [--status s] [--account-id a]   查看发布记录
+  log-get <log_id>                        查看单条发布记录(JSON)
+  health | version
+"""
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import io
+import json
+import os
+import sqlite3
+import subprocess
+import sys
+import time
+from datetime import datetime
+from typing import Any, Dict, List, Optional
+
+import requests
+from playwright.async_api import async_playwright
+
+if sys.platform == "win32":
+    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
+    sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding="utf-8", errors="replace")
+
+SKILL_SLUG = "sohu-publisher"
+SKILL_VERSION = "1.2.0"
+BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+OPENCLAW_DIR = os.path.dirname(BASE_DIR)
+
+
+def _now_unix() -> int:
+    return int(time.time())
+
+
+def _unix_to_iso(ts: Optional[int]) -> Optional[str]:
+    if ts is None:
+        return None
+    try:
+        return datetime.fromtimestamp(int(ts)).isoformat(timespec="seconds")
+    except (ValueError, OSError, OverflowError):
+        return None
+
+
+def get_data_root() -> str:
+    env = (os.getenv("CLAW_DATA_ROOT") or os.getenv("JIANGCHANG_DATA_ROOT") or "").strip()
+    if env:
+        return env
+    if sys.platform == "win32":
+        return r"D:\claw-data"
+    return os.path.join(os.path.expanduser("~"), ".claw-data")
+
+
+def get_user_id() -> str:
+    return (os.getenv("CLAW_USER_ID") or os.getenv("JIANGCHANG_USER_ID") or "").strip() or "_anon"
+
+
+def get_skills_root() -> str:
+    env = (os.getenv("CLAW_SKILLS_ROOT") or os.getenv("JIANGCHANG_SKILLS_ROOT") or "").strip()
+    if env:
+        return env
+    return OPENCLAW_DIR
+
+
+def get_skill_data_dir() -> str:
+    path = os.path.join(get_data_root(), get_user_id(), SKILL_SLUG)
+    os.makedirs(path, exist_ok=True)
+    return path
+
+
+def get_db_path() -> str:
+    return os.path.join(get_skill_data_dir(), "sohu-publisher.db")
+
+
+def get_conn() -> sqlite3.Connection:
+    return sqlite3.connect(get_db_path())
+
+
+def init_db() -> None:
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            """
+            CREATE TABLE IF NOT EXISTS publish_logs (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                account_id TEXT NOT NULL,
+                article_id INTEGER NOT NULL,
+                article_title TEXT,
+                status TEXT NOT NULL,
+                error_msg TEXT,
+                created_at INTEGER NOT NULL,
+                updated_at INTEGER NOT NULL
+            )
+            """
+        )
+        cur.execute("PRAGMA table_info(publish_logs)")
+        cols = {r[1] for r in cur.fetchall()}
+        if "article_title" not in cols:
+            cur.execute("ALTER TABLE publish_logs ADD COLUMN article_title TEXT")
+        if "updated_at" not in cols:
+            cur.execute("ALTER TABLE publish_logs ADD COLUMN updated_at INTEGER")
+        if "created_at" not in cols:
+            cur.execute("ALTER TABLE publish_logs ADD COLUMN created_at INTEGER")
+        conn.commit()
+    finally:
+        conn.close()
+
+
+def save_publish_log(account_id: str, article_id: int, article_title: str, status: str, error_msg: Optional[str] = None) -> int:
+    init_db()
+    now = _now_unix()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            """
+            INSERT INTO publish_logs (account_id, article_id, article_title, status, error_msg, created_at, updated_at)
+            VALUES (?, ?, ?, ?, ?, ?, ?)
+            """,
+            (account_id, int(article_id), article_title or "", status, error_msg, now, now),
+        )
+        new_id = int(cur.lastrowid)
+        conn.commit()
+    finally:
+        conn.close()
+    return new_id
+
+
+def check_entitlement(skill_slug: str) -> tuple[bool, str]:
+    auth_base = (os.getenv("JIANGCHANG_AUTH_BASE_URL") or "").strip().rstrip("/")
+    if not auth_base:
+        return True, ""
+    user_id = (os.getenv("CLAW_USER_ID") or os.getenv("JIANGCHANG_USER_ID") or "").strip()
+    if not user_id:
+        return False, "鉴权失败：缺少用户身份（CLAW_USER_ID / JIANGCHANG_USER_ID）"
+
+    auth_api_key = (os.getenv("JIANGCHANG_AUTH_API_KEY") or "").strip()
+    timeout = int((os.getenv("JIANGCHANG_AUTH_TIMEOUT_SECONDS") or "5").strip())
+    headers = {"Content-Type": "application/json"}
+    if auth_api_key:
+        headers["Authorization"] = f"Bearer {auth_api_key}"
+    payload = {
+        "user_id": user_id,
+        "skill_slug": skill_slug,
+        "trace_id": (os.getenv("JIANGCHANG_TRACE_ID") or "").strip(),
+        "context": {"entry": "main.py"},
+    }
+    try:
+        res = requests.post(f"{auth_base}/api/entitlements/check", json=payload, headers=headers, timeout=timeout)
+    except requests.RequestException as exc:
+        return False, f"鉴权请求失败：{exc}"
+    if res.status_code != 200:
+        return False, f"鉴权服务异常：HTTP {res.status_code}"
+    try:
+        body = res.json()
+    except ValueError:
+        return False, "鉴权服务异常：返回非 JSON"
+    code = body.get("code")
+    data = body.get("data") or {}
+    if code != 200:
+        return False, str(body.get("msg") or "鉴权失败")
+    if not data.get("allow", False):
+        return False, str(data.get("reason") or "未购买或已过期")
+    return True, ""
+
+
+def _call_json_script(script_path: str, args: List[str]) -> Optional[Dict[str, Any]]:
+    proc = subprocess.run([sys.executable, script_path, *args], capture_output=True, text=True, encoding="utf-8", errors="replace")
+    raw = (proc.stdout or "").strip()
+    if not raw or raw.startswith("ERROR"):
+        return None
+    try:
+        return json.loads(raw)
+    except json.JSONDecodeError:
+        return None
+
+
+def _normalize_sohu_title(raw_title: str, article: Dict[str, Any]) -> str:
+    """
+    搜狐标题约束：5-72 字。
+    - >72：截断
+    - <5：用正文首行/前缀补足
+    """
+    title = (raw_title or "").strip()
+    if len(title) > 72:
+        title = title[:72].rstrip()
+
+    if len(title) >= 5:
+        return title
+
+    body = str(article.get("content") or article.get("content_html") or "").strip()
+    first_line = ""
+    for line in body.splitlines():
+        t = line.strip()
+        if t:
+            first_line = t
+            break
+    seed = first_line or "搜狐发布稿件"
+    seed = seed.replace("\r", " ").replace("\n", " ").strip()
+
+    # 先把 seed 拼上，仍不足时再补固定后缀，最终保证 >=5
+    merged = (title + seed).strip()
+    if len(merged) < 5:
+        merged = (merged + "发布稿件标题").strip()
+    if len(merged) > 72:
+        merged = merged[:72].rstrip()
+    return merged
+
+
+def get_account(account_id: str) -> Optional[Dict[str, Any]]:
+    script = os.path.join(get_skills_root(), "account-manager", "scripts", "main.py")
+    return _call_json_script(script, ["get", str(account_id)])
+
+
+def get_article(article_id: str) -> Optional[Dict[str, Any]]:
+    script = os.path.join(get_skills_root(), "content-manager", "scripts", "main.py")
+    return _call_json_script(script, ["get", str(article_id)])
+
+
+async def publish(account: Dict[str, Any], article: Dict[str, Any], account_id: str) -> str:
+    profile_dir = account["profile_dir"]
+    original_title = str(article.get("title") or "")
+    title = _normalize_sohu_title(original_title, article)
+    if title != original_title:
+        print(f"ℹ️ 标题已自动修正（搜狐要求 5-72 字）：\n  原标题：{original_title}\n  新标题：{title}")
+    content_html = article.get("content_html", article.get("content", ""))
+
+    async with async_playwright() as p:
+        browser = await p.chromium.launch_persistent_context(
+            user_data_dir=profile_dir,
+            headless=False,
+            channel="chrome",
+            no_viewport=True,
+            permissions=["clipboard-read", "clipboard-write"],
+            args=["--start-maximized"],
+        )
+        page = browser.pages[0] if browser.pages else await browser.new_page()
+        await page.goto("https://mp.sohu.com/mpfe/v4/contentManagement/news/addarticle?contentStatus=1")
+        await page.wait_for_load_state("networkidle")
+
+        try:
+            title_input = page.locator(".publish-title input").first
+            await title_input.wait_for(state="visible", timeout=10000)
+        except Exception:
+            await browser.close()
+            return "ERROR:REQUIRE_LOGIN"
+
+        print("💡 页面加载且已确认登录，开始自动填入文字...")
+        await title_input.click()
+        await title_input.fill("")
+        await page.keyboard.insert_text(title)
+        await asyncio.sleep(1)
+
+        editor = page.locator("#editor .ql-editor").first
+        await editor.click()
+        await page.evaluate(
+            """(html_str) => {
+                const blobHtml = new Blob([html_str], { type: 'text/html' });
+                const blobText = new Blob([html_str], { type: 'text/plain' });
+                const item = new window.ClipboardItem({
+                    'text/html': blobHtml,
+                    'text/plain': blobText
+                });
+                return navigator.clipboard.write([item]);
+            }""",
+            content_html,
+        )
+
+        modifier = "Meta" if sys.platform == "darwin" else "Control"
+        await page.keyboard.press(f"{modifier}+v")
+        await asyncio.sleep(3)
+        await page.locator("li.publish-report-btn").first.click()
+        print("⌛ 正在提交发布，进入高压视觉核验阶段...")
+
+        publish_success = False
+        error_text = "动作执行阻断：由于特殊元素拦截、频率限制或底层报错，未能成功发出。"
+        try:
+            async with page.expect_navigation(url=lambda u: "addarticle" not in u, timeout=8000):
+                pass
+            publish_success = True
+        except Exception:
+            try:
+                limit_text = page.locator("text=/.*已达上限.*/").first
+                if await limit_text.is_visible(timeout=1500):
+                    error_text = await limit_text.inner_text()
+                else:
+                    error_msg = await page.evaluate(
+                        """() => {
+                            const els = Array.from(document.querySelectorAll('div, span, p'));
+                            for (let el of els) {
+                                const style = window.getComputedStyle(el);
+                                if ((style.position === 'fixed' || style.position === 'absolute')
+                                    && parseInt(style.zIndex || 0) > 80
+                                    && el.innerText.trim().length > 3
+                                    && el.innerText.trim().length < 80) {
+                                    return el.innerText.trim();
+                                }
+                            }
+                            return null;
+                        }"""
+                    )
+                    if error_msg:
+                        error_text = f"抓取到报错原文: {error_msg}"
+            except Exception as e:
+                error_text = f"抓取报错文案期间遭遇环境隔离: {e}"
+
+        await asyncio.sleep(5)
+        await browser.close()
+        if publish_success:
+            return "SUCCESS"
+        return f"FAIL:{error_text}"
+
+
+def cmd_publish(account_id: str, article_id: str) -> int:
+    ok, reason = check_entitlement(SKILL_SLUG)
+    if not ok:
+        print(f"❌ {reason}")
+        return 1
+    if not str(article_id).isdigit():
+        print("❌ article_id 必须是数字。请先执行 content-manager 的 list 查看 id。")
+        return 1
+
+    account = get_account(account_id)
+    if not account:
+        print(f"❌ 查无此配置账号：{account_id}")
+        return 1
+    platform = str(account.get("platform") or "").strip().lower()
+    if platform != "sohu":
+        platform_cn_map = {
+            "doubao": "豆包",
+            "deepseek": "DeepSeek",
+            "qianwen": "通义千问",
+            "kimi": "Kimi",
+            "yiyan": "文心一言",
+            "yuanbao": "腾讯元宝",
+            "toutiao": "头条号",
+            "zhihu": "知乎",
+            "wechat": "微信公众号",
+            "sohu": "搜狐号",
+        }
+        got_cn = platform_cn_map.get(platform, platform or "未知平台")
+        print("❌ 账号平台不匹配：当前账号不是「搜狐号」。")
+        print(f"当前 account_id={account_id} 对应平台：{got_cn}（platform={platform or 'unknown'}）")
+        print("请换一个搜狐账号 id 后重试。")
+        print("可先执行：python account-manager/scripts/main.py list sohu")
+        return 1
+    login_status = int(account.get("login_status") or 0)
+    if login_status != 1:
+        print("❌ 该搜狐账号当前未登录，暂不能发布。")
+        print("请先手工登录，再执行发布：")
+        print(f"  python account-manager/scripts/main.py login {account_id}")
+        print(f"登录完成后再执行：python sohu-publisher/scripts/main.py publish {account_id} {article_id}")
+        return 1
+    article = get_article(article_id)
+    if not article:
+        print(f"❌ 查无此文章编号（库中无 ID: {article_id}）")
+        return 1
+
+    result = asyncio.run(publish(account, article, account_id))
+    content_script = os.path.join(get_skills_root(), "content-manager", "scripts", "main.py")
+    title = article.get("title", "")
+
+    if result == "ERROR:REQUIRE_LOGIN":
+        save_publish_log(account_id, int(article_id), title, "require_login", "账号未登录或登录已失效")
+        print(f"⚠️ 搜狐号 ({account_id}) 登录状态已失效，发布流程已中止。")
+        print("请先手工完成登录，再重新发布：")
+        print(f"  python account-manager/scripts/main.py login {account_id}")
+        print(f"  python sohu-publisher/scripts/main.py publish {account_id} {article_id}")
+        return 1
+
+    if result == "SUCCESS":
+        log_id = save_publish_log(account_id, int(article_id), title, "published", None)
+        subprocess.run([sys.executable, content_script, "feedback", article_id, "published", account_id])
+        print(f"🎉 发布成功：{title}")
+        print(f"✅ 发布日志已记录，log_id={log_id}")
+        return 0
+
+    if result.startswith("FAIL:"):
+        error_msg = result[len("FAIL:") :]
+        log_id = save_publish_log(account_id, int(article_id), title, "failed", error_msg)
+        subprocess.run([sys.executable, content_script, "feedback", article_id, "failed", account_id, error_msg])
+        print(f"❌ 发布失败：{error_msg}")
+        print(f"✅ 失败日志已记录，log_id={log_id}")
+        return 1
+
+    save_publish_log(account_id, int(article_id), title, "failed", f"未知结果：{result}")
+    return 1
+
+
+def cmd_logs(limit: int = 10, status: Optional[str] = None, account_id: Optional[str] = None) -> int:
+    init_db()
+    if limit <= 0:
+        limit = 10
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        sql = (
+            "SELECT id, account_id, article_id, article_title, status, error_msg, created_at, updated_at "
+            "FROM publish_logs WHERE 1=1 "
+        )
+        params: List[Any] = []
+        if status:
+            sql += "AND status = ? "
+            params.append(status)
+        if account_id:
+            sql += "AND account_id = ? "
+            params.append(account_id)
+        sql += "ORDER BY created_at DESC, id DESC LIMIT ?"
+        params.append(int(limit))
+        cur.execute(sql, tuple(params))
+        rows = cur.fetchall()
+    finally:
+        conn.close()
+
+    if not rows:
+        print("暂无发布记录")
+        return 0
+
+    sep_line = "_" * 39
+    for idx, r in enumerate(rows):
+        rid, aid, arid, title, st, err, cat, uat = r
+        print(f"id：{rid}")
+        print(f"account_id：{aid or ''}")
+        print(f"article_id：{arid}")
+        print(f"article_title：{title or ''}")
+        print(f"status：{st or ''}")
+        print(f"error_msg：{err or ''}")
+        print(f"created_at：{_unix_to_iso(cat) or str(cat or '')}")
+        print(f"updated_at：{_unix_to_iso(uat) or str(uat or '')}")
+        if idx != len(rows) - 1:
+            print(sep_line)
+            print()
+    return 0
+
+
+def cmd_log_get(log_id: str) -> int:
+    if not str(log_id).isdigit():
+        print("❌ log_id 必须是数字")
+        return 1
+    init_db()
+    conn = get_conn()
+    try:
+        cur = conn.cursor()
+        cur.execute(
+            "SELECT id, account_id, article_id, article_title, status, error_msg, created_at, updated_at FROM publish_logs WHERE id = ?",
+            (int(log_id),),
+        )
+        row = cur.fetchone()
+    finally:
+        conn.close()
+    if not row:
+        print("❌ 没有这条发布记录")
+        return 1
+    rid, aid, arid, title, st, err, cat, uat = row
+    print(
+        json.dumps(
+            {
+                "id": int(rid),
+                "account_id": aid,
+                "article_id": int(arid),
+                "article_title": title,
+                "status": st,
+                "error_msg": err,
+                "created_at": _unix_to_iso(cat),
+                "updated_at": _unix_to_iso(uat),
+            },
+            ensure_ascii=False,
+        )
+    )
+    return 0
+
+
+class ZhArgumentParser(argparse.ArgumentParser):
+    def error(self, message: str) -> None:
+        print(f"参数错误：{message}\n请执行：python main.py -h 查看帮助", file=sys.stderr)
+        self.exit(2)
+
+
+def _print_full_usage() -> None:
+    print("搜狐号发布（main.py）可用命令：")
+    print("  python main.py publish <account_id> <article_id>   # 发布一篇")
+    print("  python main.py logs [--limit N] [--status s] [--account-id a]   # 查看发布记录")
+    print("  python main.py log-get <log_id>                    # 查看单条日志(JSON)")
+    print("  python main.py health")
+    print("  python main.py version")
+    print()
+    print("常见示例：")
+    print("  python main.py publish sohu_account1 12")
+    print("  python main.py logs")
+    print("  python main.py logs --status failed --limit 20")
+    print("  python main.py log-get 7")
+    print()
+    print("说明：也兼容旧写法：python main.py <account_id> <article_id>")
+
+
+def build_parser() -> ZhArgumentParser:
+    p = ZhArgumentParser(
+        prog="main.py",
+        description="搜狐号发布：发布文章、查看发布记录、查询单条日志。",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+        epilog=(
+            "示例：\n"
+            "  python main.py publish sohu_account1 12\n"
+            "  python main.py logs\n"
+            "  python main.py logs --status failed --limit 20\n"
+            "  python main.py log-get 7\n"
+            "  python main.py health\n"
+            "  python main.py version"
+        ),
+    )
+    sub = p.add_subparsers(dest="cmd", required=True, parser_class=ZhArgumentParser)
+
+    sp = sub.add_parser("publish", help="发布一篇文章到搜狐号")
+    sp.add_argument("account_id", help="账号 id（来自 account-manager list）")
+    sp.add_argument("article_id", help="文章 id（来自 content-manager list）")
+    sp.set_defaults(handler=lambda a: cmd_publish(a.account_id, a.article_id))
+
+    sp = sub.add_parser("logs", help="查看发布记录（默认最近 10 条）")
+    sp.add_argument("--limit", type=int, default=10, help="最多显示条数（默认 10）")
+    sp.add_argument("--status", default=None, help="按状态筛选：published/failed/require_login")
+    sp.add_argument("--account-id", default=None, help="按账号 id 筛选")
+    sp.set_defaults(handler=lambda a: cmd_logs(limit=a.limit, status=a.status, account_id=a.account_id))
+
+    sp = sub.add_parser("log-get", help="按 log_id 查看单条发布记录(JSON)")
+    sp.add_argument("log_id", help="日志 id（整数）")
+    sp.set_defaults(handler=lambda a: cmd_log_get(a.log_id))
+
+    sp = sub.add_parser("health", help="健康检查")
+    sp.set_defaults(handler=lambda _a: 0 if sys.version_info >= (3, 10) else 1)
+
+    sp = sub.add_parser("version", help="版本信息(JSON)")
+    sp.set_defaults(
+        handler=lambda _a: (
+            print(json.dumps({"version": SKILL_VERSION, "skill": SKILL_SLUG}, ensure_ascii=False)) or 0
+        )
+    )
+    return p
+
+
+def main(argv: Optional[List[str]] = None) -> int:
+    argv = argv if argv is not None else sys.argv[1:]
+    if not argv:
+        _print_full_usage()
+        return 1
+    # 兼容旧用法：python main.py <account_id> <article_id>
+    if len(argv) == 2 and argv[0] not in {"publish", "logs", "log-get", "health", "version", "-h", "--help"}:
+        return cmd_publish(argv[0], argv[1])
+    parser = build_parser()
+    args = parser.parse_args(argv)
+    return int(args.handler(args))
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

toutiao-publisher/.github/workflows/release_skill.yaml

@@ -0,0 +1,13 @@
+# 打 tag（如 v1.0.2）后触发：加密打包、调用平台 API 同步元数据（入库）、上传制品、清理旧版本。
+# 逻辑在 admin/jiangchang-platform-kit 的 reusable-release-skill.yaml 中。
+name: 头条技能发布
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  release:
+    uses: admin/jiangchang-platform-kit/.github/workflows/reusable-release-skill.yaml@main
+    with:
+      artifact_platform: windows
+      pyarmor_platform: windows.x86_64

toutiao-publisher/.gitignore

@@ -0,0 +1,8 @@
+__pycache__/
+*.py[cod]
+*.pyo
+.Python
+.venv/
+venv/
+.env
+.env.*

toutiao-publisher/README.md

@@ -0,0 +1,40 @@
+# 头条号批量发布（toutiao-publisher）
+
+头条号批量发布技能仓库；当前为**占位阶段**，仅含 CLI 骨架与文档，**发布功能待后续迭代**。
+
+## 目录一览
+
+| 路径 | 作用 |
+|------|------|
+| `SKILL.md` | 技能清单（YAML 头 + Markdown 正文） |
+| `release.ps1` | 本地一键发布（依赖与 `jiangchang-platform-kit` 并列，见 `account-manager` 同款） |
+| `scripts/skill_main.py` | CLI 入口：`health` / `version` |
+| `docs/` | 运行时与可移植性说明 |
+| `optional/` | 可选片段（路径、SQLite 等），默认不引用 |
+
+## 本地试跑
+
+```bash
+python scripts/skill_main.py health
+python scripts/skill_main.py version
+```
+
+## 版本
+
+与 `SKILL.md` 中 `version` 字段对齐更新。
+
+## Git 远程（避免推到模板仓）
+
+从 `skill-template` 克隆的目录，默认 `origin` 仍指向模板仓库。本技能应使用独立仓库，例如：
+
+```text
+http://120.25.191.12:3000/admin/toutiao-publisher.git
+```
+
+在 Gitea 上**新建同名空仓库**后执行：
+
+```bash
+git remote set-url origin http://120.25.191.12:3000/admin/toutiao-publisher.git
+git push -u origin main
+git push origin v1.0.1   # 若本地已有 tag 且需同步
+```

toutiao-publisher/SKILL.md

@@ -0,0 +1,52 @@
+---
+# ---------------------------------------------------------------------------
+# 技能清单（Skill Manifest）
+# ---------------------------------------------------------------------------
+name: 头条号批量发布
+description: 头条号批量发布技能（骨架阶段：仅健康检查与版本；发布逻辑待实现）。
+version: 0.1.0
+author: 深圳匠厂科技有限公司
+metadata:
+  openclaw:
+    slug: toutiao-publisher
+    emoji: "📰"
+    category: "内容发布"
+  skill:
+    slug: toutiao-publisher
+    emoji: "📰"
+    category: "内容发布"
+allowed-tools:
+  - bash
+---
+
+# 头条号批量发布（toutiao-publisher）
+
+## 使用时机
+
+- 用户需要**在头条号侧批量或自动化发布内容**时（具体话术与流程待业务实现后补充）。
+
+## 执行步骤
+
+### 健康检查
+
+```bash
+python3 {baseDir}/scripts/skill_main.py health
+```
+
+### 查看版本
+
+```bash
+python3 {baseDir}/scripts/skill_main.py version
+```
+
+### 发布与其它子命令
+
+实现中：后续将在 `scripts/` 下增加发布入口，并在此文档补充命令与参数说明。
+
+## 环境依赖
+
+详见本仓库 `docs/RUNTIME.md`（`CLAW_DATA_ROOT`、`CLAW_USER_ID` 等）。
+
+## 数据与隐私
+
+本技能若产生持久化数据，应仅写入 `{CLAW_DATA_ROOT}/{CLAW_USER_ID}/toutiao-publisher/` 下；不得将用户数据提交到版本库。

toutiao-publisher/docs/PORTABILITY.md

@@ -0,0 +1,56 @@
+# 多宿主（多 Claw）可移植性说明
+
+## 设计目标
+
+同一套技能仓库应能在**不同 Claw 实现**下工作，只要宿主满足本文档的**最小契约**：能启动进程、传入环境变量、并把 `SKILL.md` 提供给编排层阅读。
+
+## 行业上常见的「技能包」形态
+
+- **声明式清单**（Markdown + YAML 头）：描述名称、描述、版本、工具权限、触发场景。
+- **可执行入口**：一至多个脚本/二进制，由宿主通过 `bash` / `python` 等调用。
+- **用户数据与代码分离**：持久化数据落在用户可写目录，不写在安装目录内。
+
+本仓库按上述惯例组织；具体宿主如何解析 `SKILL.md` 的 YAML 键名，以各宿主文档为准。
+
+## 标识技能：`metadata.skill`
+
+`SKILL.md` 中使用 **`metadata.skill.slug`**（及 `metadata.openclaw.slug`）作为**可移植**的机器可读标识（短横线命名，如 `toutiao-publisher`）。
+
+若你的宿主仍要求其它键名（例如历史实现里的嵌套字段），请在宿主侧做**映射**，或在 `SKILL.md` 中**并列声明**两组 metadata（保持 `slug` 值一致）。不要在业务代码里写死某一宿主品牌名。
+
+## 环境变量：推荐前缀 `CLAW_*`
+
+为减少对单一产品名的耦合，文档与可选片段推荐使用：
+
+| 变量 | 含义 |
+|------|------|
+| `CLAW_DATA_ROOT` | 用户数据根目录（多技能共享的上一级） |
+| `CLAW_USER_ID` | 当前工作空间或用户标识，用于数据隔离 |
+| `CLAW_SKILLS_ROOT` | 可选；多个技能并排安装时的根目录，便于 `subprocess` 调用兄弟技能 |
+
+宿主若已使用其它名称，推荐在**启动子进程时**注入别名，例如：
+
+- 将宿主内部的「数据根」映射为 `CLAW_DATA_ROOT`
+- 将宿主内部「用户 ID」映射为 `CLAW_USER_ID`
+
+这样技能脚本无需分支判断宿主品牌。
+
+## 路径布局约定（逻辑路径）
+
+在 `CLAW_DATA_ROOT` 与 `CLAW_USER_ID` 可用时，本技能推荐将私有数据放在：
+
+```text
+{CLAW_DATA_ROOT}/{CLAW_USER_ID}/{skill_slug}/
+```
+
+其中 `skill_slug` 与 `SKILL.md` 中 `metadata.skill.slug` 一致。若环境变量缺失，`optional/paths_snippet.py` 中提供了**仅用于开发机**的 fallback（见该文件注释），生产环境应由宿主注入变量。
+
+## 发布与制品
+
+不同组织对加密、签名、制品格式要求不同。`.github/workflows/release_skill.yaml` 仅作占位：**务必替换**为你们自己的复用工作流或删除。
+
+## 自检
+
+- [ ] `SKILL.md` 中 `slug` 与目录名/制品名策略是否与宿主一致  
+- [ ] 宿主文档要求的环境变量是否已全部注入  
+- [ ] 是否在文档中说明了「未注入变量时的行为」（拒绝运行 / 本地 fallback）

toutiao-publisher/docs/RUNTIME.md

@@ -0,0 +1,55 @@
+# 运行时契约（环境变量与目录）
+
+本文档定义技能进程**建议依赖**的外部条件，便于不同 Claw 宿主统一接入。业务技能应**读取环境变量**，而不是在代码里写死路径或用户名。
+
+## 必需程度说明
+
+- **强烈建议**：生产环境由宿主注入；技能应对缺失给出明确错误提示，避免静默写到意外目录。
+- **可选**：没有时技能仍可部分运行（例如只读 `health`）。
+
+## 变量一览
+
+### `CLAW_DATA_ROOT`（强烈建议）
+
+用户数据根。多个技能、多个用户的数据都在此根之下分区。
+
+- 典型场景：组织策略指定的盘符路径或 `~/.your-org-data`。
+- 未设置时：技能**不应**猜测网络盘；开发机 fallback 仅限 `optional/paths_snippet.py` 中说明的情形。
+
+### `CLAW_USER_ID`（强烈建议）
+
+当前会话所代表的用户或工作空间 ID（字符串）。与数据隔离强相关。
+
+- 用于拼接：`{CLAW_DATA_ROOT}/{CLAW_USER_ID}/{skill_slug}/`
+- 未设置时：可用匿名占位（如 `_anon`）**仅用于开发**，生产应显式注入。
+
+### `CLAW_SKILLS_ROOT`（可选）
+
+多个技能以并列目录安装时的根路径，例如：
+
+```text
+{CLAW_SKILLS_ROOT}/skill-a/scripts/...
+{CLAW_SKILLS_ROOT}/skill-b/scripts/...
+```
+
+编排型技能若需要通过子进程调用兄弟技能，应基于该变量定位脚本，避免写死绝对路径。
+
+## 本技能推荐的数据目录
+
+```text
+{CLAW_DATA_ROOT}/{CLAW_USER_ID}/{skill_slug}/
+```
+
+- `skill_slug`：与 `SKILL.md` 内 `metadata.skill.slug` 一致。
+- 在此目录下可放置 SQLite 文件、缓存、上传临时文件等；**不要**向版本库提交该目录内容。
+
+## 标准输出约定（建议）
+
+为便于宿主与自动化解析，建议：
+
+- 致命错误：单行前缀 `ERROR:`，例如 `ERROR:MISSING_ENV_CLAW_DATA_ROOT`
+- 成功：人类可读一行或多行；若有机读需求，可用 `JSON` 单行输出并在 `SKILL.md` 中说明格式。
+
+## 与具体宿主的关系
+
+若某宿主文档规定了另一套变量名，应在**宿主启动技能子进程时**注入为本文档中的 `CLAW_*` 名称，或在技能内使用一层极薄的 `getenv` 封装（见 `optional/paths_snippet.py` 注释示例）。**不要在业务模块中散落多套变量名判断。**

toutiao-publisher/docs/SKILL_TYPES.md

@@ -0,0 +1,33 @@
+# 技能形态与自检清单（无业务）
+
+开发前先选定形态，避免把「编排、存储、浏览器」混在一个脚本里难以测试。
+
+## 类型 A：无状态工具型
+
+- **特征**：不持久化用户数据，或只读配置文件；输入输出主要在 stdin/stdout。
+- **数据目录**：通常不需要 `CLAW_DATA_ROOT` 下的专属库；若需要缓存，仍建议放在契约目录下。
+- **自检**：离线可跑；`health` 不访问网络也可成功。
+
+## 类型 B：本地持久化型
+
+- **特征**：使用 SQLite、本地文件等保存用户数据。
+- **数据目录**：必须使用 `{CLAW_DATA_ROOT}/{CLAW_USER_ID}/{skill_slug}/`。
+- **自检**：首次运行自动建库/建表；文档中说明库文件路径与备份方式。
+
+## 类型 C：编排型（调用其它技能或外部 CLI）
+
+- **特征**：自身逻辑薄，主要 `subprocess` 或 HTTP 调用其它组件。
+- **依赖**：在 `SKILL.md` 中明确写出**先决条件**（兄弟技能已安装、某 CLI 在 PATH 等）。
+- **自检**：`health` 可检查兄弟可执行文件是否存在；缺失时打印清晰错误。
+
+## 类型 D：混合型
+
+- **特征**：既有本地存储，又调用外部能力。
+- **建议**：拆模块（存储 / 编排 / 领域逻辑），入口脚本只做参数解析与调度。
+
+## 发布前通用自检
+
+- [ ] `SKILL.md` 中触发条件与示例命令与实际入口一致  
+- [ ] 未注入 `CLAW_DATA_ROOT` / `CLAW_USER_ID` 时行为已文档化  
+- [ ] 不向仓库提交用户数据、密钥、大型二进制  
+- [ ] 错误信息包含「如何修复」（缺什么环境变量、缺哪个依赖）

toutiao-publisher/optional/README.md

@@ -0,0 +1,17 @@
+# optional/ 目录说明
+
+本目录下的文件**不会**被 `scripts/skill_main.py` 自动引用。
+
+## 为什么要单独放
+
+- 避免可选片段拖慢最小 `health` 起步，按需再复制进业务代码。
+- 需要时**整文件复制**到 `scripts/` 或你们自己的包路径下，再按文件头注释改名、改常量。
+
+## 文件列表
+
+| 文件 | 用途 |
+|------|------|
+| `paths_snippet.py` | `CLAW_*` 数据目录解析与 fallback 说明 |
+| `sqlite_minimal.py` | 无业务含义的 SQLite 建表示例 |
+
+本技能数据子目录与 `SKILL_SLUG` 为 `toutiao-publisher`；若复制片段到其它项目请改为对应 slug 与表名。

toutiao-publisher/optional/paths_snippet.py

@@ -0,0 +1,92 @@
+# -*- coding: utf-8 -*-
+"""
+可选片段：路径与环境变量（请复制到 scripts/ 或作为独立模块使用）
+================================================================
+
+【用途】
+  统一解析「用户数据根」「用户 ID」「本技能私有目录」，避免在业务代码里重复 os.getenv。
+
+【使用步骤】
+  1. 将本文件复制到 skills/your-slug/scripts/paths.py（或任意模块名）。
+  2. 把 SKILL_SLUG 改为与 SKILL.md 中 metadata.skill.slug 一致。
+  3. 在业务代码中: from paths import get_skill_data_dir  （按实际包路径调整）
+
+【可移植性】
+  - 优先读取标准名 CLAW_*（见 docs/RUNTIME.md）。
+  - 若你的组织在宿主侧仍使用历史变量名，可在此文件 _aliases 列表中追加 (标准名, 备选名)，
+    由宿主注入其一即可；不要在业务里再写第三套名字。
+
+【注意】
+  - 未设置 CLAW_DATA_ROOT 时的 fallback 仅适合开发机；生产环境应由宿主注入。
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+
+# TODO: 复制本文件后改为你的 slug（与 SKILL.md metadata.skill.slug 一致）
+SKILL_SLUG = "toutiao-publisher"
+
+
+def _getenv_first(names: tuple[str, ...]) -> str:
+    """按顺序读取多个环境变量名，返回第一个非空值。"""
+    for n in names:
+        v = (os.getenv(n) or "").strip()
+        if v:
+            return v
+    return ""
+
+
+def get_data_root() -> str:
+    """
+    用户数据根目录。
+    顺序：CLAW_DATA_ROOT → （可选）宿主别名，见下方元组。
+    若皆空：Windows 默认 D:\\claw-data；其它系统默认 ~/.claw-data —— 仅开发便利，生产请注入 CLAW_DATA_ROOT。
+    """
+    root = _getenv_first(
+        (
+            "CLAW_DATA_ROOT",
+            # 在此追加组织内别名，例如 "MYORG_USER_DATA_ROOT",
+        )
+    )
+    if root:
+        return root
+    if sys.platform == "win32":
+        return r"D:\claw-data"
+    return os.path.join(os.path.expanduser("~"), ".claw-data")
+
+
+def get_user_id() -> str:
+    """当前用户或工作空间 ID；未设置时用 _anon（仅开发）。"""
+    uid = _getenv_first(
+        (
+            "CLAW_USER_ID",
+            # 在此追加别名，例如 "MYORG_WORKSPACE_ID",
+        )
+    )
+    return uid or "_anon"
+
+
+def get_skill_data_dir() -> str:
+    """
+    本技能可写目录：{数据根}/{用户ID}/{skill_slug}/
+    会自动 os.makedirs(..., exist_ok=True)。
+    """
+    path = os.path.join(get_data_root(), get_user_id(), SKILL_SLUG)
+    os.makedirs(path, exist_ok=True)
+    return path
+
+
+def get_skills_root() -> str:
+    """
+    可选：并列安装的多技能根目录。
+    未设置 CLAW_SKILLS_ROOT 时，默认使用本文件所在仓库的上一级目录的上一级
+    （即：.../toutiao-publisher/optional/ → 仅作开发时并列技能推断参考）。
+    """
+    root = _getenv_first(("CLAW_SKILLS_ROOT",))
+    if root:
+        return root
+    # optional/ 下：仓库根为 dirname(dirname(__file__))
+    here = os.path.dirname(os.path.abspath(__file__))
+    return os.path.dirname(here)

toutiao-publisher/optional/sqlite_minimal.py

@@ -0,0 +1,48 @@
+# -*- coding: utf-8 -*-
+"""
+可选片段：最小 SQLite 示例（请复制后按需修改）
+==============================================
+
+【用途】
+  演示「单表 + 自增主键 + INTEGER 时间戳」的一种严谨写法；与具体业务无关。
+
+【使用步骤】
+  1. 复制到 scripts/db_example.py（或并入你的模块）。
+  2. 修改 TABLE_SQL 中的表名与字段；保持时间字段为 INTEGER Unix 秒（UTC）若需跨时区一致。
+  3. 在入口脚本中仅在需要持久化时 import。
+
+【注意】
+  - 本示例不做迁移兼容；schema 变更请用你们组织的迁移策略。
+  - 数据库文件路径建议：get_skill_data_dir() / "skill.db"（paths_snippet 中函数）。
+"""
+
+from __future__ import annotations
+
+import sqlite3
+import time
+
+# 示例表：与任何业务无关
+TABLE_SQL = """
+CREATE TABLE IF NOT EXISTS skill_audit (
+    id INTEGER PRIMARY KEY AUTOINCREMENT, -- 自增主键
+    action TEXT NOT NULL,                -- 动作标识，如 health_ok
+    created_at INTEGER NOT NULL          -- Unix 秒 UTC
+);
+"""
+
+
+def connect(db_path: str) -> sqlite3.Connection:
+    return sqlite3.connect(db_path)
+
+
+def init_db(conn: sqlite3.Connection) -> None:
+    conn.execute(TABLE_SQL)
+    conn.commit()
+
+
+def record_action(conn: sqlite3.Connection, action: str) -> None:
+    conn.execute(
+        "INSERT INTO skill_audit (action, created_at) VALUES (?, ?)",
+        (action, int(time.time())),
+    )
+    conn.commit()

toutiao-publisher/release.ps1

@@ -0,0 +1,23 @@
+[CmdletBinding()]
+param(
+  [string]$Prefix = "v",
+  [string]$Message = "正式发布",
+  [switch]$AutoCommit,
+  [switch]$RequireClean,
+  [string]$CommitMessage,
+  [switch]$DryRun
+)
+
+Set-StrictMode -Version Latest
+$ErrorActionPreference = "Stop"
+
+$scriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+$sharedScript = Join-Path $scriptDir "..\jiangchang-platform-kit\tools\release.ps1"
+$sharedScript = [System.IO.Path]::GetFullPath($sharedScript)
+
+if (-not (Test-Path $sharedScript)) {
+  throw "Shared release script not found: $sharedScript"
+}
+
+& $sharedScript @PSBoundParameters
+exit $LASTEXITCODE

toutiao-publisher/scripts/skill_main.py

@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+头条号发布技能 — CLI 入口
+========================
+
+【职责】
+  - 作为宿主调用时的统一 CLI 入口（子进程、终端、CI 均可）。
+  - 只做：参数解析、环境检查、分发到具体子命令；复杂逻辑放到同目录其它模块。
+
+【如何扩展】
+  1. 在 main() 的 dispatch 字典中增加 "your_cmd": handler 项。
+  2. 实现 handler(argv) 或 handler()；出错时打印 ERROR: 前缀信息并 sys.exit(非0)。
+  3. 在仓库根目录 SKILL.md「执行步骤」中补充示例命令。
+
+【多宿主注意】
+  - 不要在本文件写死某一品牌宿主名。
+  - 路径与环境变量约定见 ../docs/RUNTIME.md；可选辅助代码见 ../optional/paths_snippet.py（需自行复制或 import 路径按项目调整）。
+
+【编码】
+  Windows 下若宿主仍使用系统默认编码，可在宿主侧设置 UTF-8；此处不强制改 sys.stdout（避免与宿主捕获冲突）。
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from typing import Callable, Dict, List, Optional
+
+# 与 SKILL.md 中 metadata.openclaw.slug / metadata.skill.slug 保持一致
+SKILL_SLUG = "toutiao-publisher"
+
+
+def cmd_version(_args: argparse.Namespace) -> int:
+    """打印版本信息（与 SKILL.md frontmatter 中 version 应对齐，此处为占位）。"""
+    payload = {
+        "skill_slug": SKILL_SLUG,
+        "version": "0.1.0",
+        "entry": "skill_main.py",
+    }
+    print(json.dumps(payload, ensure_ascii=False))
+    return 0
+
+
+def cmd_health(_args: argparse.Namespace) -> int:
+    """
+    健康检查：应快速、可离线（除非技能本身强依赖网络）。
+    失败时打印 ERROR: 前缀，便于宿主与自动化解析。
+    """
+    # 示例：检查 Python 版本（可按需改为检查关键依赖 import）
+    if sys.version_info < (3, 9):
+        print("ERROR:PYTHON_VERSION need >= 3.9", file=sys.stderr)
+        return 1
+    print(f"OK skill={SKILL_SLUG} python={sys.version.split()[0]}")
+    return 0
+
+
+def build_parser() -> argparse.ArgumentParser:
+    p = argparse.ArgumentParser(
+        description="toutiao-publisher — Toutiao batch publish skill CLI (skeleton).",
+    )
+    sub = p.add_subparsers(dest="command", required=True)
+
+    sp = sub.add_parser("version", help="Print version JSON.")
+    sp.set_defaults(handler=cmd_version)
+
+    sp = sub.add_parser("health", help="Quick health check.")
+    sp.set_defaults(handler=cmd_health)
+
+    return p
+
+
+def main(argv: Optional[List[str]] = None) -> int:
+    argv = argv if argv is not None else sys.argv[1:]
+    parser = build_parser()
+    args = parser.parse_args(argv)
+    handler: Callable[[argparse.Namespace], int] = args.handler
+    return handler(args)
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())