cloak-mcp demonstrates solid definition quality with well-structured tool schemas and clear security-focused descriptions. All 6 tools have explicit descriptions (avg 156 chars, within 10-1024 baseline) and comprehensive input schemas with typed parameters. Naming follows verb_noun convention (list_, get_, sign_, proxy_, mint_, query_). However, output schemas are not documented in the visible source, and error handling guidance is minimal. Parameter descriptions are present but could be more prescriptive about constraints and recovery paths. The security-centric design (never exposing secrets, redacting auth headers) is exemplary, but lacks explicit error classification and recovery guidance patterns.

• Document output schemas for all 6 tools. Specify field names, types, and structure. Example: list_secret_names returns {secrets: [{name: string, kind: string, tags: string[]}], total: integer}.
• Add error classification and recovery guidance to each tool description. Example: 'If secret_name not found, returns error code SECRET_NOT_FOUND. Call list_secret_names() to discover available secrets.'
• Add JSON Schema constraints for numeric parameters: duration_seconds (min 1, max 3600), limit (min 1, max 1000), offset (min 0). Document header validation rules for sign_request and proxy_authenticated_http_request.
• For write operations (proxy_authenticated_http_request, mint_short_lived_token), add idempotency guarantees or dry-run support. Document whether repeated calls with identical parameters produce identical results or side effects.
• Enhance query_audit description with pagination details: 'Returns up to limit entries. If total > limit + offset, more results exist. Use offset to fetch next page.' Include total_count in response schema.