Anchor has 6 tools with explicit schemas and descriptions. Naming is clear and verb-based (search_, verify_, cite, safe_edit, restore_). Descriptions are present but brief (avg 80 chars, below the 194-char baseline). Input schemas are properly typed with Zod. Critical gaps: no output schemas documented in code, no error handling guidance, no parameter constraints (enums, ranges), and write operations (safe_edit, restore_note) lack dry-run/confirmation patterns despite claiming them in descriptions. The ping tool is trivial but well-formed. Overall, definitions are functional but lack LLM-optimization depth.

• Document output schemas for all tools. E.g., search_notes should declare it returns [{note_id, title, content, score, path}]. This enables LLMs to chain tools correctly.
• Expand tool descriptions to 100-200 chars. Add WHEN to use (e.g., 'Use verify_grounding to fact-check claims before citing them') and dependencies (e.g., 'Requires vault to be indexed').
• Add parameter constraints: search_notes.limit should be 1-100 (default 20); query should be 1-500 chars. Use Zod refinements or JSON Schema constraints.
• Expose the dry-run flow for safe_edit as a two-step pattern: first call returns a preview with a confirmation token, second call with token executes. Or add a 'dryRun' boolean parameter.
• Add error handling guidance in descriptions: 'Returns error if claim is not grounded (score < threshold). Suggest calling search_notes to find supporting evidence.' This guides LLM recovery.

Inferred effective spec: 2025-06-18+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry