SE
search
>-
Install
mkdir -p .claude/skills/search-v1ceversaa && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/13612" && unzip -o skill.zip -d .claude/skills/search-v1ceversaa && rm skill.zipInstalls to .claude/skills/search-v1ceversaa
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Use when the user wants to find existing content in the repo — "有没有关于 X 的笔记"、"我之前读过哪些 Y"、"把和 Z 相关的内容列出来"、"find notes on"、"search for". This is a read-only skill: it never writes files, never modifies overviews, never updates indexes. Use even when the user does not say /search.278 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
About this skill
/search
When To Use
- The user wants to retrieve or discover existing notes, topics, syntheses, conferences, or local ideas.
- The user is asking "what do we already have" before deciding what to read or synthesize.
Hard Boundaries (Read-Only)
- Does not write any markdown files
- Does not modify
overview.md - Does not update indexes
- Does not append to
ops/log.md - Does not archive PDFs
Contrast with similar skills:
/search: find and list existing content, explain relevance/synthesize: compare content and produce a new written output/discuss: use retrieved content to push a research idea forward
Workflow
- Extract key terms from the user's query.
- Run
query_context.pywith the query:
Addpython3 .codex/skills/maintain/scripts/query_context.py \ "<user query terms>" \ --limit 8--include-privateif the user may be looking for local ideas or drafts. - Present results in ranked order. For each result, provide:
- Path and title
- 1–2 sentences explaining why it's relevant to the query
- Current
statusif informative (e.g.,draftingvscomplete)
- Recommend a next step for the most relevant results:
- Note is complete and the user wants to review →
/refresh - Note is missing and user wants to read the paper →
/reador/intake - Multiple relevant notes suggest a comparison →
/synthesize - Results spark a research idea →
/discuss
- Note is complete and the user wants to review →
Inputs Expected
- Free-text query (topic, paper title fragment, concept, author name, venue)
Scripts
../maintain/scripts/query_context.py— full invocation example in_shared/script-registry.md
Done Criteria
- Ranked results are presented with relevance explanations
- A next step is suggested for the most relevant hits