Mixed quality across 12 tools. Strengths: all tools have descriptions (10 - 200 chars, within baseline); most parameters have type definitions and descriptions; tool names follow verb_noun convention (search_*, list_*, get_*, write_*). Weaknesses: several tools lack input schemas in visible code (calculate, write_markdown); no documented output schemas for any tool; no error handling guidance; no pagination support for list_documents; parameter descriptions are minimal (avg ~40 chars vs baseline 72); no enum constraints on free-form inputs; no idempotency or confirmation patterns for write operations (remember, write_markdown).

• Document output schemas for all 12 tools. Include field names, types, and whether results are paginated. Example: search_documents should return {query: string, matches: [{citation: string, excerpt: string}], note: string}.
• Add input schemas to calculate and write_markdown. Show parameter types, constraints (e.g., expression must be valid Python math syntax; filename must be 1-255 chars, no path separators).
• Implement pagination for list_documents: add limit (default 20, max 100) and offset parameters; return total count and next_offset in response.
• Add error handling guidance to write operations. E.g., remember should return {stored: string, id: string} on success, or {error: string, suggestion: string} on failure (e.g., 'Fact too long. Max 1000 chars. Try splitting into multiple facts.').
• Expand parameter descriptions to 50-100 chars. Include format hints: 'Folder path (absolute or relative to home, e.g., ~/Downloads or /tmp/data)' instead of just 'Folder path to measure'.