autocrew 0.1.0

package/skills/platform-rewrite/SKILL.md

@@ -0,0 +1,125 @@
+---
+name: platform-rewrite
+description: |
+  平台 native 改写 skill。把一个母稿改成不同平台的版本，强调"重写而不是裁剪"。支持单平台和多平台一键改写，自动生成各平台标题和 hashtag。
+---
+
+# 平台改写
+
+## 目标
+
+把一篇母稿改写成各平台原生版本。重写，不是裁剪。
+
+支持平台：小红书 · 抖音 · 公众号 · 视频号 · B站
+
+## 规则
+
+0. **When adapting titles across platforms, do NOT just truncate. Re-craft using `skills/title-craft/SKILL.md` methodology fitted to the target platform's character limits and audience behavior.**
+1. Never call a simple trim "adaptation".
+2. Start from the strongest single angle in the source draft.
+3. Generate the first structured platform version via `autocrew_rewrite`.
+4. If the user asks for multiple platforms, use the **batch flow** (Step 2 below).
+5. Run `humanizer-zh` before final delivery.
+
+## 第一步 — 单平台改写
+
+From an existing draft:
+
+```json
+{
+  "action": "adapt_platform",
+  "content_id": "content-xxx",
+  "target_platform": "douyin",
+  "save_as_draft": true
+}
+```
+
+From raw text:
+
+```json
+{
+  "action": "adapt_platform",
+  "title": "原始标题",
+  "body": "原始正文",
+  "target_platform": "xiaohongshu"
+}
+```
+
+After rewrite completes:
+
+1. **Generate title variants + hashtags** for the target platform:
+   - Use `title-hashtag.ts` → `generateForPlatform(topic, platform, { tags })`
+   - Pick the best title variant, show all alternatives to user.
+   - Attach hashtags to the saved draft via `autocrew_content` action="update":
+     ```json
+     { "action": "update", "id": "<new-content-id>", "hashtags": ["#tag1", "#tag2"] }
+     ```
+
+2. **Run humanizer** on the adapted text if it reads too smooth or generic.
+
+3. **Show output** to user: adapted title, body, hashtags, and notes.
+
+## 第二步 — 多平台批量改写
+
+When user says "帮我改成所有平台" or specifies 2+ platforms:
+
+1. Determine target platforms. Default all 5 if user says "全平台":
+   `["xiaohongshu", "douyin", "wechat_mp", "wechat_video", "bilibili"]`
+
+2. **Generate titles + hashtags for all platforms in one call:**
+   - Use `title-hashtag.ts` → `generateForAllPlatforms(baseTopic, platforms, { tags })`
+   - This returns `PlatformTitleResult[]` with titles, hashtags, and tips per platform.
+
+3. **Rewrite each platform** sequentially via `autocrew_rewrite`:
+   ```
+   for each platform:
+     1. autocrew_rewrite action="adapt_platform" → get adapted body
+     2. autocrew_content action="save" → save as new draft with platform-specific title + hashtags
+     3. autocrew_humanize (if needed)
+   ```
+
+4. **Link siblings**: after all variants are saved, update each with sibling IDs:
+   ```json
+   { "action": "update", "id": "<content-id>", "siblings": ["id-1", "id-2", "id-3"] }
+   ```
+
+5. **Summary table** — show user a comparison:
+
+   | 平台 | 标题 | 字数 | Hashtags | Content ID |
+   |------|------|------|----------|------------|
+   | 小红书 | ... | 800 | 8 | content-xxx |
+   | 抖音 | ... | 600 | 5 | content-yyy |
+   | ... | ... | ... | ... | ... |
+
+## 第三步 — 改写后审核（可选）
+
+If `creator-profile.json` has `styleCalibrated: true`:
+
+- Auto-run `autocrew_review` action="full_review" on each adapted version.
+- Flag any version that fails review.
+- Offer to auto-fix or let user decide.
+
+## 输出
+
+Always tell the user:
+
+- Which platform version(s) you created
+- Whether each was saved as a new draft (with content ID)
+- Title alternatives from title-hashtag
+- Hashtags attached
+- What still needs manual polish
+- Sibling relationships (if batch)
+
+## 错误处理
+
+| 故障 | 处理 |
+|------|------|
+| 源内容未找到 | 向用户索要 content_id 或原始文本 |
+| 不支持的平台 | 列出支持平台，让用户选择 |
+| title-hashtag 返回空 | 回退到原始标题 + tags 基础标签 |
+| 保存失败 | 在聊天中输出改写文本让用户复制，重试一次 |
+
+## 变更日志
+
+- 2026-04-01: v2 — Added multi-platform batch flow, title-hashtag integration, sibling linking, optional post-rewrite review.
+- 2026-03-31: v1 — Initial single-platform rewrite skill.

package/skills/pre-publish/SKILL.md

@@ -0,0 +1,142 @@
+---
+name: pre-publish
+description: |
+  发布前检查清单 skill。在内容发布前自动检查所有前置条件是否满足，输出 checklist 状态。只有全部通过才允许发布。
+---
+
+# 发布前检查
+
+> 门控技能。在发布前拦截，确保内容质量和完整性。
+
+## 触发时机
+
+- User says "发布", "推送", "publish", "上线"
+- User runs `autocrew_publish`
+- Agent is about to transition content to `publishing` status
+
+## 检查项
+
+对目标内容运行以下检查，每项 pass/fail：
+
+### 1. Content Review — 审核通过
+
+```json
+{ "action": "full_review", "content_id": "<id>", "platform": "<platform>" }
+```
+
+- ✅ Pass: `passed === true` (no sensitive words, no AI traces, quality ≥ 60)
+- ❌ Fail: show `summary` and `fixes`, offer auto-fix
+
+### 2. Cover Review — 封面审核通过 (XHS/Douyin only)
+
+```json
+{ "action": "read", "content_id": "<id>" }
+```
+
+Check `cover-review.json` in the content's asset directory:
+- ✅ Pass: `status === "approved"` and `approvedLabel` exists
+- ⏭️ Skip: platform is `wechat_mp` or `bilibili` (cover optional)
+- ❌ Fail: no cover review or status is `pending`/`reviewing`
+
+### 3. Hashtags — 标签已生成
+
+Check content metadata:
+- ✅ Pass: `hashtags` array exists and has ≥ 1 item
+- ❌ Fail: no hashtags → offer to generate via `title-hashtag.ts`
+
+### 4. Title — 标题符合平台规范
+
+Check title against platform rules from `title-hashtag.ts`:
+- ✅ Pass: title length within `titleLengthRange` for the platform
+- ⚠️ Warn: title exists but exceeds `maxTitleLength` → suggest trimming
+- ❌ Fail: no title
+
+### 5. Platform Set — 平台已指定
+
+- ✅ Pass: `platform` field is set and is a supported value
+- ❌ Fail: platform is empty or "unset"
+
+### 6. Body Length — 正文字数达标
+
+| Platform | Min chars |
+|----------|-----------|
+| xiaohongshu | 200 |
+| douyin | 100 |
+| wechat_mp | 800 |
+| wechat_video | 100 |
+| bilibili | 200 |
+
+- ✅ Pass: body length ≥ platform minimum
+- ❌ Fail: too short
+
+## 执行流程
+
+```
+1. Load content via autocrew_content action="get"
+2. Run checks 1-6
+3. Build checklist output
+4. If ALL pass → allow publish, transition to "publish_ready"
+5. If ANY fail → block publish, show what needs fixing
+```
+
+## 输出格式
+
+<output_template lang="zh-CN">
+```
+📋 发布前检查 — content-xxx (小红书)
+
+✅ 内容审核：通过 (质量 78/100)
+✅ 封面审核：已选定 A 方案
+✅ Hashtags：8 个标签
+✅ 标题规范：「标题内容」(16字，符合 10-18 范围)
+✅ 平台设置：xiaohongshu
+✅ 正文字数：856 字 (≥200)
+
+🟢 全部通过，可以发布！
+```
+</output_template>
+
+Or if something fails:
+
+<output_template lang="zh-CN">
+```
+📋 发布前检查 — content-yyy (抖音)
+
+✅ 内容审核：通过
+❌ 封面审核：未完成 → 运行 autocrew_cover_review 创建候选
+✅ Hashtags：5 个标签
+⚠️ 标题规范：「标题太长了超过限制」(32字，超出 25 上限)
+✅ 平台设置：douyin
+✅ 正文字数：420 字
+
+🔴 2 项未通过，请先修复再发布。
+```
+</output_template>
+
+## 自动修复建议
+
+检查失败时，提供具体修复方案：
+
+| Failed Check | Suggested Action |
+|-------------|-----------------|
+| Content review | `autocrew_review` action="auto_fix" |
+| Cover review | `autocrew_cover_review` action="create_candidates" |
+| No hashtags | `generateHashtags(topic, platform, tags)` then update content |
+| Title too long | `generateTitleVariants(topic, platform)` then pick shorter one |
+| No platform | Ask user which platform |
+| Body too short | Suggest expanding with more examples/data points |
+
+## 与发布流程的集成
+
+If all checks pass and user confirms:
+
+1. Transition content to `publish_ready`:
+   ```json
+   { "action": "transition", "id": "<id>", "target_status": "publish_ready" }
+   ```
+
+2. Then proceed with the actual publish action (e.g. `autocrew_publish`).
+
+## 变更日志
+
+- 2026-04-01: v1 — Initial pre-publish checklist with 6 checks, auto-fix suggestions, and publish gate.