Librarian has 12 well-named tools with consistent verb_noun patterns (librarian_search, librarian_add, librarian_delete). All tools have descriptions (avg ~80 chars, within baseline 34-392). Input schemas are present and typed for all tools. However, descriptions are terse and lack LLM-optimized guidance on WHEN to use each tool vs alternatives. Parameter descriptions are minimal (e.g., 'Entry title' vs 'A short, human-readable name for the activity (max 200 chars)'). Output schemas are not documented, agents cannot predict response structure. Error handling is absent from tool definitions. Write tools require session_label but lack confirmation/dry-run patterns for destructive operations (delete). No tool annotations (readOnlyHint, destructiveHint) despite clear risk stratification in metadata.

• Add output schema documentation to all tools. Example: librarian_search returns {entries: [{id, title, description, date, tags}], total_count, has_more}. Agents need to know what fields to extract.
• Expand tool descriptions to 100-150 chars with WHEN guidance. Example: 'Search entries by keyword. Use this for broad queries; use librarian_similar to find near-duplicates before creating new entries.'
• Add tool annotations: readOnlyHint=true for read tools (search, get, list, tags, schema, stats); destructiveHint=true for librarian_delete; idempotentHint=true for idempotent writes.
• Enhance parameter descriptions with constraints. Example: 'Entry title (required, 1-200 chars, no leading/trailing whitespace)' instead of 'Entry title'.
• Implement confirmation pattern for librarian_delete: add optional dry_run=true parameter or require explicit confirmation via a separate confirm_delete tool.