Critical gaps across all dimensions. Only 2 of 12 tools have visible input schemas in the provided source (fingerprint, create_job, get_job_status, list_jobs from mcp_server.py). Tools 5-12 are listed in Dockerfile without schema definitions visible, cannot verify their actual registration or parameter types. Descriptions are present but generic (avg ~80 chars, below 194-char baseline). No parameter descriptions for most tools. No output schemas documented. No error handling guidance. No tool annotations. STDIO transport caps protocol readiness at 50.

• Add complete input schemas for tools 5-12 (get_application_map, expand_application_node, etc.). Include type, description, and constraints for every parameter. Example: 'ownership_scope' should document valid enum values and when to use each.
• Document output schemas for all 12 tools. Specify returned fields, types, and which IDs/references are needed for chaining to other tools. Example: get_application_map should document the structure of nodes and edges returned.
• Add parameter descriptions to all tools. For 'operation' in create_job, list valid operations and when to use each. For 'limit'/'offset' in list_jobs, specify ranges (e.g., limit 1-100, default 20).
• Implement error handling with recovery guidance. Example: 'If artifact not found, call list_jobs() to see available artifacts' or 'If job_id is invalid (12 hex chars required), check the job ID format.'
• Add tool annotations to the MCP protocol layer. Mark read-only tools with readOnlyHint=true, destructive tools (create_job) with destructiveHint=true, and idempotent tools accordingly. This enables agents to reason about safety.