
Weread Skills
Let your coding agent search WeChat Reading, browse your shelf, export highlights, and pull reading stats while you research topics or capture book notes.
Overview
weread-skills is an agent skill for the Idea phase that queries WeChat Reading via the Agent API Gateway for search, shelf, notes, reviews, stats, and book recommendations.
Install
npx skills add https://github.com/tencent/wechatreading --skill weread-skillsWhat is this skill?
- Eight documented capabilities: search, book info, shelf, reading stats, personal notes, chapter hot highlights, public r
- Single Agent API Gateway (POST i.weread.qq.com/api/agent/gateway) with api_name routing and mandatory skill_version in e
- Bearer WEREAD_API_KEY (wrk-*) auth with vid injected automatically for user-scoped endpoints
- Per-capability reference docs (search.md, book.md, shelf.md, readdata.md, notes.md, review.md, discover.md) for paramete
- Few-shot guidance: business params flat at JSON body top level alongside api_name
- 8 capability areas documented in the skills table (search through discover)
- Mandatory skill_version 1.0.3 on every gateway request
- Unified POST entry at i.weread.qq.com/api/agent/gateway
Adoption & trust: 11.8k installs on skills.sh; 47 GitHub stars; 2/3 security scanners passed (skills.sh audits).
What problem does it solve?
You want book search, highlights, and reading analytics from WeRead without leaving your agent chat or guessing gateway parameters.
Who is it for?
Builders who already use 微信读书 and want agent-driven lookup, note export, and discover flows during research or personal knowledge capture.
Skip if: Teams that do not use WeChat Reading, need offline EPUB editing, or want English-only bookstore APIs without a wrk- API key.
When should I use this skill?
User asks to search 微信读书, check shelf, view notes/划线, reading stats, book reviews, or get recommendations via WeRead.
What do I get? / Deliverables
The agent returns structured WeRead data (books, notes, stats, reviews) using documented api_name calls and skill_version 1.0.3.
- Gateway JSON responses for the requested WeRead capability (search results, shelf, notes, stats, reviews, discover lists
Recommended Skills
Journey fit
Solo builders often read on WeRead to validate ideas and gather references; this skill sits on the research shelf as the canonical place to query books, notes, and discovery feeds before committing to a build. Research subphase covers competitor and domain learning—search, notes export, chapter hot highlights, and personalized recommendations are direct inputs to that work.
How it compares
Use instead of generic web search when your sources and notes already live in WeRead’s ecosystem.
Common Questions / FAQ
Who is weread-skills for?
Solo and indie builders (and small teams) who use WeChat Reading and want their coding agent to search books, manage shelf context, and read back notes and stats through the official agent gateway.
When should I use weread-skills?
Use it during Idea research when you need store search, reading progress, exported highlights, review sentiment, or personalized recommendations; it also helps Grow content workflows when you summarize what you read this month.
Is weread-skills safe to install?
It calls Tencent’s WeRead gateway with your Bearer API key, so treat WEREAD_API_KEY as a secret. Review the Security Audits panel on this Prism page and rotate keys if you share the skill across machines.
SKILL.md
READMESKILL.md - Weread Skills
# WeRead — 微信读书助手 通过 Agent API Gateway 调用微信读书接口,提供搜索、书架、笔记、书评等能力。 ## 支持的能力 | 能力 | 说明 | 用户示例 | 详细说明 | |------|------|----------|----------| | 搜索书籍 | 在书城搜索 | "帮我搜一下三体" | `search.md` | | 书籍信息 | 查看书籍详情、章节目录、阅读进度 | "这本书有多少章" "我读到哪了" | `book.md` | | 书架管理 | 查看书架 | "看看我的书架" | `shelf.md` | | 阅读统计 | 阅读时长、天数、偏好分析、阅读统计摘要 | "我这个月读了多久" "今年读了几本书" | `readdata.md` | | 笔记划线 | 查看个人笔记数量与内容,包括划线、想法/点评、书签数量 | "看看我在三体里的笔记" "导出我的划线" "在这本书有多少笔记" | `notes.md` | | 章节热门划线 | 查看书籍/章节热门划线、划线热度及划线下想法 | "看看这章有什么热门划线" "这段话下面有什么想法" | `notes.md` | | 书籍点评 | 查看书籍的公开点评 | "三体这本书有什么点评?" "看看推荐的点评" | `review.md` | | 推荐好书 | 个性化推荐/相似推荐 | "给我推荐几本书" | `discover.md` | 根据用户意图参考对应说明文件了解接口参数、回包结构和工作流。 --- ## 接口调用规范 ### 统一入口 ``` POST https://i.weread.qq.com/api/agent/gateway ``` ### 鉴权 - Header:`Authorization: Bearer $WEREAD_API_KEY` - `WEREAD_API_KEY` 从环境变量获取,格式 `wrk-xxxxxxxx` - 若未设置,提示用户:`export WEREAD_API_KEY=<你的apikey>` - API Key 绑定用户身份(vid),需要用户身份的接口会自动注入,无需手动传 vid ### 请求格式 - **Method**:POST - **Content-Type**:application/json - **Body**:JSON,`api_name` 指定接口,其余为接口参数,**每次请求必须带 `skill_version`** ```bash curl -X POST "https://i.weread.qq.com/api/agent/gateway" \ -H "Authorization: Bearer $WEREAD_API_KEY" \ -H "Content-Type: application/json" \ -d '{"api_name": "/store/search", "keyword": "三体", "count": 10, "skill_version": "1.0.3"}' ``` ### 请求 few-shot **正确:业务参数平铺在 body 顶层。** ```json {"api_name":"/user/notebooks","count":100,"skill_version":"1.0.5"} ``` **正确:下一页继续平铺 `lastSort`。** ```json {"api_name":"/user/notebooks","count":100,"lastSort":1516907353,"skill_version":"1.0.5"} ``` **错误:不要把业务参数包在 `params` 内。** ```json {"api_name":"/user/notebooks","params":{"count":100,"lastSort":1516907353},"skill_version":"1.0.5"} ``` 上面的错误写法会导致 `count`、`lastSort` 未被转发,后端按默认值返回第一页,看起来像分页失效。 ### 响应格式 - JSON,回包经过字段裁剪,只返回核心字段 - `errcode` 非 0 时表示错误,给出中文提示 - 发送 `{"api_name": "/_list"}` 可查看所有可用接口及参数定义 ### 通用规则 1. **版本上报**:每次请求 body 必须包含 `"skill_version": "1.0.3"`(取本文件顶部 version 字段的值),用于服务端检查版本更新。**如果回包中出现 `upgrade_info` 字段,必须立即暂停当前操作,按照 `upgrade_info.message` 中的指引完成升级,升级完成后再重新执行用户请求,不得忽略该字段** 2. **参数平铺**:业务参数必须和 `api_name`、`skill_version` 放在同一层;不要包在 `params`、`data`、`body` 等对象里。只有接口文档明确声明的数组/对象字段(如 `/book/readreviews` 的 `reviews`)才允许作为业务字段传入。 3. **能力文档预检**:调用任何接口前,必须先根据「支持的能力」表阅读对应说明文件(如阅读统计先读 `readdata.md`,书架先读 `shelf.md`),确认接口参数、字段含义、单位、计数口径和工作流;禁止仅凭字段名或经验猜测含义。 4. **字段解释优先级**:解释接口回包时,以对应说明文件中的字段说明为准;如果回包字段名和直觉含义冲突,必须服从说明文件,不得直接翻译字段名。 5. **bookId 解析**:用户输入书名时,先调 `/store/search` 获取 bookId,再执行后续操作 6. **书架数量**:使用 `/shelf/sync` 回答“书架有多少本书/多少条目”时,必须按 `books.length + albums.length + (mp 非空 ? 1 : 0)` 计算;`albums[]` 是专辑/有声书,也属于书架里的书,详细规则见 `shelf.md` 7. **结果展示**:列表用编号展示方便选择;搜索结果重点展示书名、作者、评分;展示接口回包信息时,字段**禁止**直接翻译,应该参考文件中的说明内容提供 8. **上下文衔接**:对话中记住已查询的 bookId,后续操作无需用户重复提供 9. **深度链接**:在展示划线、想法、章节等内容时,拼接对应的跳转链接方便用户直接在 App 中打开,具体格式见下方「深度链接(URL Schema)」章节 10. **数据展示规范**: - **时间戳**:所有 Unix 时间戳字段(如 `updateTime`、`createTime`、`finishTime`、`readUpdateTime` 等),**展示时须转为 YYYY-MM-DD 格式**(如 `1748563200` 展示为"2025-05-30"),不得直接展示原始数字 - **阅读时长**:单位为秒,展示时转为"X小时Y分钟"格式 --- ## 深度链接(URL Schema) 在展示书籍、章节、划线等内容时,如果回包字段足以构造链接,应附上对应的跳转链接,方便用户点击后直接在微信读书 App 中打开对应位置。想法/点评不一定都有划线位置,只有具备 `chapterUid` 和 `range` 时才生成划线位置链接。 ### 打开书籍(跳转到上次阅读进度) ``` weread://reading?bId={bookId} ``` | 参数 | 说明 | 来源 | |------|------|------| | `bookId` | 书籍 ID | 各接口返回的 `bookId` | **示例**: ``` weread://reading?bId=3300045871 ``` **使用场景**: - 展示书架列表时,每本书附上跳转链接 - 展示搜索结果时,附上「打开阅读」链接 - 展示阅读进度时,提供「继续阅读」链接 ### 跳转到指定章节 ``` weread://reading?bId={bookId}&chapterUid={chapterUid} ``` | 参数 | 说明 | 来源 | |------|------|------| | `bookId` | 书籍 ID | 各接口返回的 `bookId` | | `chapterUid` | 章节 UID | `/book/chapterinfo` 返回的 `chapters[].chapterUid` | **示例**: ``` weread://reading?bId=3300045871&chapterUid=107 ``` **使用场景**: - 展示章节目录时,每个章节附上跳转链接 ### 跳转到划线/想法所在位置 ``` weread://bestbookmark?bookId={bookId}&chapterUid={chapterUid}&rangeStart={rangeS