Single tool 'gate_check' has a well-structured schema with proper enums and descriptions. The tool description is comprehensive (194 chars) and explains the four checks, verdict types, and when to call it. Input parameters are properly typed with enum constraints for 'only'/'skip' and boolean for 'strict'. However, the tool lacks output schema documentation, the response structure (pass|warn|fail verdict, blocking reasons) is described in prose but not formally specified. No error handling guidance is provided. The tool is read-only (no state mutation), which is appropriate.

• Document the output schema explicitly: { verdict: 'pass'|'warn'|'fail', blocking_reasons: [{check: string, reason: string}], warnings: [...], metadata: {duration_ms: number, checks_run: string[]} }. Include field types and descriptions.
• Add error handling guidance: 'If gate engine is not installed, return error with recovery hint: "Install @nugehs/gate: npm i -g @nugehs/gate or npm i @nugehs/gate in your project."'
• Clarify parameter interactions: 'If both only and skip are provided, skip takes precedence. If only is empty, all checks run (unless skip excludes them).'
• Rename tool to 'check_ship_readiness' or 'verify_deployment_gates' to follow verb_noun convention and improve LLM intent recognition.
• Add a 'format' parameter (enum: 'summary'|'detailed') to let agents control response verbosity and keep token usage predictable.