メニュー
Plan discipline, risk scaling, verification, lessons, and bounded subagent/delegation rules for this repo.
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
Canonical rules — feature layers, domain purity, cubit vs widget roles, DI entrypoints, Freezed. Part of agents-canonical-rules split.
Scaffold or add a feature Cubit with Freezed state, tests, and repo-standard placement under presentation/cubit/.
Figma design-to-code for this repo (figma-sync assets, Flutter adaptation, MCP).
Dependency direction and feature composition. Use for shared package/app-shell changes, cross-feature deps, auth/theme contracts.
External references and key file paths for this repo. Use when locating theme, DI, lifecycle, sync, HTTP, or Supabase code.
Flutter BLoC app stack snapshot and lib/ layout. Use at task start or when adding features, routes, or shared code.
| name | agents-meta-behavior |
| description | Plan discipline, risk scaling, verification, lessons, and bounded subagent/delegation rules for this repo. |
Canon: AGENTS.md, docs/agent_knowledge_base.md, docs/ai_code_review_protocol.md, tasks/lessons.md. Task roles (multi-agent hub): docs/agent_knowledge_base.md#multi-agent-hub — Researcher/Analyst (subagent_type: explore), Implementer (subagent_type: generalPurpose), Reviewer (subagent_type: code-reviewer); untrusted; Redact secrets.
Delegation: bounded scope; main agent owns verify. context ladder → docs/ai/context_loading.md. Harness max-score: docs/ai/harness_auto_maintenance.md; ./bin/agent-maintain harness-maintain; closeout when harness scope.
