vacmcp defines 6 tools with basic descriptions and partial schemas. All tools follow verb_noun naming (search_code, trace_calls, get_code, list_contexts, compare_versions, search_history). Descriptions are present but minimal (10-60 chars), below the 50-200 char LLM-optimized baseline. Input schemas are visible for most tools with type and description fields, but output schemas are not documented. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being READ_ONLY. Error handling and recovery guidance are absent. Parameter descriptions are sparse, e.g., 'context' is described as 'The version context ID' but lacks guidance on how to discover valid context IDs or what happens if an invalid one is passed.

• Expand tool descriptions to 50-200 chars. Example: 'search_code' → 'Search for code patterns across a version context. Use this to find function definitions, imports, or specific code sequences. Returns matching file paths and line numbers.'
• Document output schemas for all tools. Specify what fields are returned (e.g., list_contexts returns [{id: string, name: string, revision: string}]) so agents can chain tools.
• Add readOnlyHint: true to all 6 tools to signal they are safe for agents to call without side effects.
• Enhance parameter descriptions with constraints and discovery hints. Example: 'context: The version context ID (e.g., "v1", "v2"). Call list_contexts() first to see available contexts.'
• Add error handling guidance to descriptions. Example: 'If context is invalid, search_code returns an error listing available contexts. If query is empty, returns all files in the context.'

Inferred effective spec: 2026-07-28+.

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