菜单
抽取 hn_story：title、url、author、points、created_at
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
搜索学术论文（Semantic Scholar + DBLP + arXiv 三源并行），返回标题、作者、摘要、引用数、链接
从 api.crossref.org 抽取学术文献：标题、链接、DOI、摘要片段及引用数
从 api.openalex.org 抽取学术作品：标题、URL、出版年份、DOI、类型
搜索微信公众号文章并提取全文（通过 Sogou 微信搜索），返回标题、真实 URL、发布时间、正文内容
搜索近期新闻文章（NewsAPI + Google News RSS 双源回退），返回标题、来源、日期、摘要
抽取 so_question：title、url、score、answer_count、tags
| name | hn_story |
| description | 抽取 hn_story：title、url、author、points、created_at |
| layer | access |
| invocation | agent_called |
| entity_type | hn_story |
| template | public_api |
| status | mined |
Search hn.algolia.com — hn_story. Content is raw JSON served inside a
tag; standard CSS selectors are ineffective, requiring JSON parsing logic.Agent-called skill — sub-agents invoke it as the typed tool
skill_hn_story(...). The dispatcher does not auto-route; the agent decides.Parameters
query (required) — The search query term.tags (default story) — Filter by tag (e.g., story, comment, poll).num_results (default 5) — Maximum number of results to return.Auto-generated by the access-skill baker (template: public_api). Hand-edit freely; the runtime only executes executor.py and reads params_schema from manifest.yaml.