This MCP server has reasonable naming conventions and decent overall structure, but suffers from inconsistent parameter documentation, missing output schema details, and lack of error handling guidance. 5 of 7 tools have descriptions exceeding the 20-char minimum, but several parameters lack descriptions entirely (e.g., 'limit', 'cursor' in query_atomics). Tool schemas are visible in the input definitions but output schemas are not documented in the rubric data provided. The server shows effort in security awareness (DESTRUCTIVE risk annotation for execute_atomic) and composition (7 single-responsibility tools), but falls short of production-grade standards in parameter clarity and error recovery guidance.

• Add description to refresh_atomics: explain it updates the local atomic tests cache from GitHub, is idempotent, and takes time. Advise LLM to call this only when tests are stale.
• Add min/max constraints to query_atomics 'limit' parameter (suggest 1-100, default 20) and format description for 'cursor' (opaque pagination token).
• Document output schema for each tool. Example for server_info: 'Returns {name: string, version: string, transport: string, os: string, data_directory: string, execution_enabled: bool}'
• Add error handling section to execute_atomic: 'Returns error if ART_EXECUTION_ENABLED != true. Call server_info to check execution_enabled status. Returns error with GUID if the test fails, use get_atomic_result(GUID) to retrieve logs.'
• Expand get_validation_schema description: 'Returns the JSON Schema that atomic test YAML must conform to. Call this before generate_atomic if you need to understand the required fields and structure.'
• Add note to generate_atomic: 'May fail if technique_id is invalid or platform is unsupported. Call query_atomics first to verify the technique exists.'
• Document validate_atomic output: what fields indicate validation success vs. which contain error details and field paths.

Inferred effective spec: <=2025-11-25.

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