Orchestrate the complete human-confirmed intent-testing workflow by using Maieutic to expose uncertainty, confirm intent, and propose missing tests (applied only on explicit request), then using Elenchus to challenge those tests with risk-directed mutations, and deliver the canonical review surface (Review This / We Verified / Still at Risk) with copy-ready inline comment candidates. Use for Feature Review or Refactor Guard of pull requests, local diffs, or AI-generated changes, or requests such as「変更の意図確認からMutation検証までやって」「人間の判断だけ聞いてテストを強化して」「AI生成PRをレビューして」「Socraticワークフローを実行して」.
Evaluate existing and changed tests by challenging them with risk-ranked mutations, measuring incremental protection and protection regressions, analyzing catches and survivors, and proving explicitly requested missing tests only in isolated workspaces. Use for standalone test-quality assessment, AI-generated test review, intent-aware catching, regression hardening, refactor guarding, critical diff validation, or requests such as「既存テストを評価して」「追加された単体テストが有効か確認して」「ミューテーションテストして」「テストがバグを検知できるか確認して」.
Analyze a code change, expose what is not yet known, draw intended behavior from repository evidence and low-friction human questions, maintain a validated Intent Contract, review unit tests with risk-appropriate QA techniques, design missing tests, run them without treating implementation as specification, and draft each justified question as a copy-ready Intent decision review comment. Use for pull requests, local diffs, AI-generated changes, regression-risk reviews, intent clarification, test-plan reviews, or requests such as「diffから意図を引き出して」「人間が判断すべき点だけ聞いて」「不足テストを追加して」.