Single tool with minimal documentation. Tool name 'limitless_query_before_work' is awkwardly long and unclear (54 chars vs baseline 18). Description is vague ('Select one permissioned exact component or source-free method, or abstain'), 95 chars, lacks WHAT/WHEN/WHY clarity. Input schema loaded from external file (query-0.1.schema.json), cannot verify parameter types, descriptions, or constraints from source. Output schema also external (decision-0.1.schema.json), not visible. Tool annotations present (readOnlyHint, openWorldHint) but insufficient to compensate for missing schema visibility. No error handling guidance in code. No parameter documentation visible.

• Rename tool to a shorter, verb-first name: 'query_component' or 'select_component' (target ~15-20 chars). Verify the name alone conveys the action.
• Rewrite description to 150-250 chars: 'Query the local component catalog to find and select a permissioned component by name or digest. Returns the component metadata and verification status. Use this before importing any external code to ensure it is trusted and immutable.'
• Embed input and output schema definitions directly in mcp_server.py (or load and validate them at startup). Ensure all parameters have type, description, and constraints (enums, patterns, min/max). Verify schema is visible in source.
• Add error handling guidance: when CatalogError is raised, return a message like 'Component not found. Available components: [list]. Try a different name or digest.' This guides LLM recovery.
• Consider adding companion tools: 'list_components' (discovery), 'verify_component' (validation), 'import_component' (execution). Single-tool servers limit agent planning and composability.