メニュー
Multi-agent orchestration enforcement for /team command. Validates session directory paths and ensures /team Tasks use a valid agent from the routing whitelist. Use when debugging /team orchestration failures.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
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。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/MadAppGang/magus --skill agent-enforcementコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Picks the browser-use agent brain LLM per task. Use when choosing a model for retry_with_browser_use_agent or a cloud session, or the user mentions a cheap, slow, or smart model.
Reference for the Browser Use MCP tools — parameters, returns, session lifecycle. Use when navigating, clicking, typing, evaluating JS, or extracting page content.
Visual UI debugging — screenshot capture and analysis, responsive layout checking at multiple viewport sizes, CSS validation via DOM state, visual regression detection, before/after state comparison.
SKILL.md を表示中
| name | agent-enforcement |
| description | Multi-agent orchestration enforcement for /team command. Validates session directory paths and ensures /team Tasks use a valid agent from the routing whitelist. Use when debugging /team orchestration failures. |
| disable-model-invocation | true |
Two-layer defense against orchestration violations in /team:
hooks/enforce-team-rules.sh) — runtime violation blockerteam.md: model: opus) — better instruction following| Model Type | Method | Tool | Reliability |
|---|---|---|---|
| Internal (Claude) | Task({resolved_agent}) | Task tool | High (same process) |
| External (Grok, Gemini, etc.) | claudish MCP tools (team/create_session) | MCP | 100% (deterministic) |
External models are called via claudish MCP tools — no Bash invocation needed.
The PreToolUse hook intercepts Task and Bash tool calls at runtime:
| Rule | Condition | Action |
|---|---|---|
| 1 | /team Task with wrong agent | DENY (must be in agent whitelist) |
| 2 | /tmp/ in Task prompt | DENY |
Detects /team workflows by vote template pattern in Task prompts. Blocks:
dev:researcher, dev:debugger, dev:developer, dev:architect, dev:test-architect, dev:devops, dev:frontend/tmp/ paths in Task prompts (use ai-docs/sessions/)The /team command uses model: opus (Opus 4.6) which follows complex XML instructions
much more reliably than Sonnet (~90% vs ~33% compliance).
| Task Type | Primary Agent | Alternatives |
|---|---|---|
| Investigation | dev:researcher | dev:debugger |
| Review | dev:reviewer | code-analysis:detective |
| Architecture | dev:architect | — |
| Implementation | dev:developer | — |
| Testing | dev:test-architect | — |
| DevOps | dev:devops | — |
| UI/Design | dev:frontend | — |
The /team command resolves the agent dynamically from the context detection table in team.md
based on task keywords (e.g., "debug" → dev:debugger, "review" → dev:researcher).
The hook enforces a whitelist of all valid agents.
Methods:
team MCP tool (handles parallel execution internally)Run the test suite to verify enforcement:
cd autotest/team && bash run-tests.sh
Hook blocking legitimate calls: The hook triggers on /team vote template patterns in Task prompts and claudish in Bash commands. Normal Task usage (without vote templates) is never affected.
resolve-agents.sh not found:
The /team command uses its own built-in logic. The hook still provides runtime protection.
Hook behavior:
The hook skips existence checks (which claudish, command -v claudish).