Infrastructure MCP has 36 tools with consistent naming (verb_noun pattern: list_, get_, run_, etc.) and basic descriptions. However, critical gaps undermine production readiness: (1) Input schemas are inferred from the tool list but NOT visible in source code, no explicit JSON Schema definitions found in the provided code samples. (2) Parameter descriptions exist but are minimal (avg ~40 chars vs baseline 72 chars). (3) Output schemas are completely undocumented, no tool returns a documented response structure. (4) Error handling is structured (toolerr package with categories, recommendations, retryable flags) but only 2 tools shown; unclear if all 36 use it. (5) No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite 10 WRITE-risk tools. (6) Secrets handling appears correct (no credentials in params, SSH/Telnet via inventory), but no audit logging visible. Average per-tool score: 58.

• Add explicit JSON Schema definitions for all 36 tools. Include type, description, enum, minLength, maxLength, pattern for every parameter. Make schemas visible in tool registration code.
• Document output schemas for every tool. Specify which fields are returned, their types, and which are safe to chain into downstream tools (e.g., pod_name from kubectl_get_pods → kubectl_logs).
• Add tool annotations: mark read-only tools with readOnlyHint=true; mark destructive tools (run_command, docker_restart, proxmox_*_vm) with destructiveHint=true; mark idempotent tools with idempotentHint=true.
• Expand parameter descriptions to 50-150 chars. Include format hints, constraints, and examples. E.g., 'target: inventory name of the target server (e.g., web-prod-01; must exist in inventory.yaml)'.
• Add pagination to list/query tools: include limit (default 20, max 100), offset/cursor, and total_count in responses. Document in tool descriptions: 'Returns up to 20 results; use offset to paginate.'

Inferred effective spec: 2025-06-18+.

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