Three meta-tools with clear naming (search_, get_, execute_) and documented schemas. However, descriptions lack depth and parameter documentation is minimal. search_capabilities and get_capability_schema have adequate descriptions (100+ chars), but execute_capability's description is terse (60 chars). No parameter descriptions visible for 'input' in execute_capability. Output schemas are not documented. Error handling and recovery guidance are absent. The design is clever (progressive disclosure pattern) but tool definitions themselves lack the rigor expected for production use.

• Add explicit output schema documentation for all three tools. Example: search_capabilities returns [{key: string, description: string, tags: [string]}]. This enables agents to extract the right fields.
• Expand execute_capability description to 150+ chars: 'Execute a capability by key with input matching its schema (obtained via get_capability_schema). Returns the capability's output. If the key is invalid, call search_capabilities() first to discover valid keys.'
• Add description to execute_capability's 'input' parameter: 'Input parameters as a JSON object. Must match the schema returned by get_capability_schema(key). Pass null or {} if the capability requires no input.'
• Document error cases for execute_capability: 'Returns error if key not found, input does not match schema, or the underlying executor fails. Check the error message for guidance.'
• Add dependency hints to get_capability_schema: 'Call search_capabilities() first to discover capability keys. Then call this tool to retrieve the full input schema before invoking execute_capability().' This guides the multi-step workflow.