This server has a single tool 'execute_skill_action' with a generic, vague description that provides no actionable guidance. The tool definition is inferred from a JSON output string rather than explicit MCP tool registration with schemas. No input parameters are documented, no output schema is defined, and no error handling guidance exists. The description (67 chars) is below the 194-char baseline for production tools and fails to explain WHAT the tool does, WHEN to use it, or what it returns. This is a non-functional MCP implementation masquerading as a skill server.

• Define the tool explicitly in MCP with a proper input schema. Include all parameters the tool accepts (e.g., collection_name, search_vector_dim, top_k) with types (string, integer) and descriptions.
• Rename the tool to reflect its actual operation: 'search_vector_partition' or 'query_distributed_index' instead of the generic 'execute_skill_action'.
• Expand the description to 100-200 chars explaining: (1) What it does: 'Search a billion-scale distributed vector index for nearest neighbors'; (2) When to use it: 'Call this to find similar embeddings in a large collection'; (3) What it returns: 'Partition ID, query node count, throughput, and latency metrics'.
• Document the output schema with all fields returned: search_partition_id (string), collection (string), vector_dimension (integer), active_query_nodes_sharded (integer), index_algorithm (string), search_throughput_qps (integer), p99_latency_ms (float), total_vectors_in_cluster (integer).
• Add error handling guidance: 'If the partition is unavailable, the tool returns an error with available alternatives. If the query times out (p99_latency_ms > 100ms), retry with a smaller top_k value.'