npx claudepluginhub telvokdev/librarianDefined in hooks/hooks.json
mcp__plugin_librarian_librarian__library_publishprintf '\n⚠️ PUBLISHING ATTESTATION REQUIRED\n\nBefore publishing, you MUST confirm with the user:\n ✓ This is original work (or they have rights to publish)\n ✓ No secrets, credentials, or sensitive data included\n ✓ Terms of service accepted\n\nDo NOT auto-confirm. Ask the user explicitly.\n'mcp__plugin_librarian_librarian__recordprintf 'New entry recorded. The semantic index will include it on next rebuild_index() call.'mcp__plugin_librarian_librarian__syncprintf 'Sync complete. Run rebuild_index() if you want new content to appear in semantic search.'mcp__plugin_librarian_librarian__import_memoriesprintf 'Import complete. Run rebuild_index() to make imported entries searchable via semantic search.'mcp__plugin_librarian_librarian__deleteprintf 'Entry deleted. Run rebuild_index() to remove stale embeddings from the search index.'mcp__plugin_librarian_librarian__library_buyprintf 'Purchase initiated. After completing payment, run sync() to download the content.'mcp__plugin_librarian_librarian__authprintf 'Auth action complete. Run auth({ action: "status" }) to verify your connection.'mcp__plugin_librarian_librarian__library_publishprintf 'Book published! Share the link or check seller_analytics() to track performance.'*printf 'Librarian active.
BRIEF: Before thinking through a problem, making decisions, or planning - check what we already know. Also use when an issue comes up again or feels like something we solved before. Call brief to search past knowledge. Use brief without a query for recent entries.
RECORD: Call when we:
- Solve a problem before moving to the next thing
- Make a decision and need to capture the reasoning NOW
- Hit a gotcha or mistake that WILL come up again
- Learn something non-obvious since context dies fast
Do not wait to be asked. Do not batch insights - one insight = one record call. Multiple recordings in one response is ok. Quality bar: I wish we knew this yesterday
MARK_HIT: When an entry from brief actually helps, call mark_hit with the path. Entries with more hits rank higher in future searches.'Runs Python scripts before 'Skill' tools (PreToolUse check), after for logging and homeostasis monitoring (PostToolUse), and aggregates daily learnings on UserPromptSubmit. No bash or file writes.