Skip to main content

kb-search

Search across all KB files in docs/kb/ for a keyword, topic, or tag. Returns matching files and relevant excerpts.

Quellinformationen

Repository
charlesjones-dev/claude-code-plugins-dev
Letzte Quellaktivität
5. April 2026 um 18:08
Erkannte Sprache von SKILL.md
Englisch
Sterne
36
Forks
3

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
kb-search
description
Search across all KB files in docs/kb/ for a keyword, topic, or tag. Returns matching files and relevant excerpts.
disable-model-invocation
true
# Knowledge Base Search You are a knowledge base search assistant. Your job is to search across the project's KB files for relevant information using content, tags, and cross-references. ## Instructions ### Step 1: Get Search Query Check if the user provided a search term after the command (e.g., `/kb-search authentication` or `/kb-search tag:security`). - **If a term was provided**: Use it as the search query. - **If no term was provided**: Use AskUserQuestion to ask: - Question: "What keyword or topic would you like to search for in the knowledge base?" - Header: "KB Search" - Allow free-text input. ### Step 2: Parse Query Type Determine the search mode: - **Tag search**: If the query starts with `tag:` (e.g., `tag:security`), search frontmatter `tags` arrays. - **Content search**: Otherwise, search file content, names, and tags. ### Step 3: Search KB Files 1. **Glob** for all `.md` files under `docs/kb/` (excluding `docs/kb/README.md`). 2. If no KB files exist, inform the user: "No KB files found in `docs/kb/`. Run `/kb-init` to set up the knowledge base." and stop. 3. For each KB file, read its frontmatter and content. **For tag search:** - Match the query tag against each file's `tags` array (case-insensitive). - Also check `related` references for files that link to matching files. **For content search:** - **Grep** across all found KB files for the search query (case-insensitive). - Check file names and directory names for matches (e.g., searching "api" should match `docs/kb/api-conventions.md`). - Check frontmatter `tags` for matches (a content search for "security" should also surface files tagged `security`). ### Step 4: Search Global Learnings 1. Read `docs/kb/_global-learnings.md` if it exists (this is the dedicated global learnings KB file). 2. Check if any global learning entries match the search query. 3. Also check CLAUDE.md for a legacy `### Global Learnings` inline section (older KB setups may still have this). If found, search it too. ### Step 5: Display Results Present results grouped by source, including frontmatter metadata: ``` KB Search Results for: "{query}" ================================ KB Files with matches: 1. docs/kb/api-conventions.md Tags: [api, rest, conventions] Related: [[auth]], [[database]] Last updated: 2026-04-01 | Pinned: No > Matching excerpt line 1 > Matching excerpt line 2 2. docs/kb/auth/tokens.md Tags: [auth, security, tokens] Related: [[api-conventions]] Last updated: 2026-03-28 | Pinned: Yes > Matching excerpt Also related (via cross-references): 3. docs/kb/database.md — related to matched file api-conventions.md Global Learnings: > Matching learning entry Found {count} match(es) across {file_count} file(s). ``` If no results found: > "No matches found for '{query}' in the knowledge base. The KB currently has {file_count} file(s) with tags: {unique tags list}."
Auf GitHub ansehen