Enforce MCP-first tool usage through a centralized gateway/bookkeeper: discover MCP capabilities, route intents to the best MCP tool, block non-MCP execution when an MCP equivalent exists, and log all tool calls/results with auditable fallbacks. Use when building or operating a multi-agent system where agents must not access tools/networks directly and all actions must be policy-enforced and recorded.
Risk-based testing, verification, and CI/CD quality gating. Use when asked to test or verify changes, create test plans, review CI/CD workflows, or assess release readiness.
Audit a repository/project for security practices, CI/CD workflow safety, adherence to project instructions (AGENTS.md/CONTRIBUTING/SECURITY), and consistency across code/config/docs. Use when preparing for release, reviewing GitHub Actions workflows, enforcing engineering guardrails, or assessing delivery readiness and governance.