メニュー
Use when splitting complex work into independent subagent tasks with clear ownership, evidence, and integration responsibilities.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | subagent-supervisor |
| description | Use when splitting complex work into independent subagent tasks with clear ownership, evidence, and integration responsibilities. |
Return the task split, each subagent's scope, accepted findings or changes, rejected findings with rationale, and final validation.
Use when adding, changing, or reviewing MCP servers, agent tools, provider integrations, credentials, scopes, or tool approval behavior.
Use when converting a product spec, use case, or PRD into a concrete implementation plan before code changes.
Use to review a diff against repository contracts, specs, docs, tests, and architecture boundaries.
Use after incidents, regressions, repeated review findings, or completed features to convert lessons into durable factory improvements.
Use when a task requires meaningful implementation. Research the repo, inspect docs/specs, and create a concrete plan before editing code.
Use when implementing behavior from an approved plan with test-first or regression-test-first development.
