菜单
Search `.agents/` durable knowledge using a local JSON index generated from markdown sources. Use `index-docs.py` to refresh search data, and optionally run docs-compile when you also want durable section index files regenerated.
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
Regenerate optional durable docs indexes and rebuild docs-search cache from local markdown sources. Use after editing `.agents/docs/` (any section), and before maintenance passes that validate index alignment.
Check the `.agents/` knowledge layer for duplication, contradiction, staleness, oversized guidance, broken or stale links, missing index coverage, uncategorized or misplaced knowledge, and consistency between `decisions/` and `troubleshooting/` indexes and their entry files. Use as a periodic maintenance pass.
Read a completed temporary session bundle from `.agents/sessions/` and distill durable repo knowledge into `.agents/` files. Use only after task-closeout has produced a bundle and when the goal is to separate stable guidance from temporary task notes.
Use whenever working with maintainer plans under .agents/docs/plans/. Covers reading a plan before implementation, creating or scaffolding a plan, and updating status or body during or after work. Trigger when the user asks to plan, scope deferred work, check what is planned, or record a plan. Also trigger at the start of implementation to see whether a plan already exists.
Capture the current task into a structured temporary session bundle under `.agents/sessions/` so a learning agent can later distill durable repo knowledge. Use for completed, blocked, or abandoned tasks with meaningful changes, debugging, validation, or reusable lessons.
Scaffold a new maintainer plan under `.agents/docs/plans/<id>.md` with valid YAML frontmatter, collision-checked ids, optional body stub sections (link decisions in a
| name | docs-search |
| description | Search `.agents/` durable knowledge using a local JSON index generated from markdown sources. Use `index-docs.py` to refresh search data, and optionally run docs-compile when you also want durable section index files regenerated. |
Search durable guidance under .agents/ with a local JSON index.
pip install pyyaml)From the target repository (or any subdirectory under it):
python3 .agents/skills/docs-search/scripts/index-docs.py
python3 .agents/skills/docs-search/scripts/search-docs.py "auth troubleshooting"
If the optional docs-compile skill is installed, you can refresh durable section indexes and search index together:
bash .agents/skills/docs-compile/scripts/docs-compile.sh
Default search output lists up to 5 matches; use --limit 3 for top-3.
index-docs.py so docs-search-index.json reflects every *.md under .agents/docs/ (plus playbooks and AGENTS.md as before).search-docs.py "<query>" to get ranked results (title, short description, path).deployment stepsauth errorsknowledge lintpython3 .agents/skills/docs-search/scripts/index-docs.py.bash .agents/skills/docs-compile/scripts/docs-compile.sh (if installed; see .agents/skills/docs-compile/SKILL.md).index-docs.py --dry-run prints sample JSON only and does not write docs-search-index.json.index-docs.py --agents-root /path/to/.agents overrides auto-discovery (equivalent to AGENTS_ROOT=/path/to/.agents).search-docs.py --agents-root /path/to/.agents matches the indexer when the skill runs from outside that repo's .agents/skills/docs-search/.search-docs.py --limit 3 returns fewer matches (default is 5).docs-search-index.json is a repo-local cache under the target checkout's .agents/skills/docs-search/ (written by index-docs.py). search-docs.py loads that path when it can resolve that checkout's .agents/ from the working directory or AGENTS_ROOT / --agents-root, so search stays scoped to the repo even if the skill definition lives elsewhere.docs-search-index.json should remain gitignored.