메뉴
Orchestrate multiple specialized subagents in a coordinated workflow. Use when running multiple agents together, delegating to agents, or executing complex multi-agent pipelines.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Remove AI-generated code smells and patterns from code. Use when cleaning up AI-generated code, removing over-commenting, simplifying verbose patterns, or making code look like a senior engineer wrote it.
Systematic code review with structured output and actionable findings. Use when reviewing code, auditing quality, checking for bugs, security vulnerabilities, or performance issues.
Create and maintain high-quality documentation using the librarian subagent. Use when writing docs, READMEs, API documentation, code summaries, or any text-based artifacts.
Design-first frontend development with modern UI/UX principles. Use when working on design, UI, styling, layout, CSS, responsive design, animations, components, or frontend visual tasks.
Atomic commits, rebase surgery, history search, and safe git operations. Use when committing, rebasing, squashing, searching git history, or performing any git operations.
Comprehensive test creation using TDD principles. Use when writing tests, improving coverage, fixing failing tests, or setting up test infrastructure.
SOC 직업 분류 기준
| name | team-run |
| description | Orchestrate multiple specialized subagents in a coordinated workflow. Use when running multiple agents together, delegating to agents, or executing complex multi-agent pipelines. |
Trigger: Use /team or when the user mentions running multiple agents together,
or asks to "use the team", "run agents", "delegate to agents".
Orchestrates multiple specialized subagents in a coordinated workflow. Each agent
has a specific role, model assignment, and tool access defined in .qwen/agents/.
| Agent | File | Role | Tools |
|---|---|---|---|
| sisyphus | sisyphus.md | Main ultraworker — decomposes & executes complex tasks | All tools |
| prometheus | prometheus.md | Strategic planner — creates phased execution plans | Read + write files, search |
| hephaestus | hephaestus.md | Deep implementation — builds features end-to-end | All tools |
| atlas | atlas.md | Architecture analyst — reviews structure & design | Read-only + search |
| explore | explore.md | Researcher — finds code, maps architecture | Read-only + search + web |
| librarian | librarian.md | Documentation specialist — writes docs, READMEs | Read + write + edit |
| code-reviewer | code-reviewer.md | Quality gate — reviews for bugs, security, perf | Read-only + search |
| testing-expert | testing-expert.md | Test specialist — writes & fixes tests | All tools |
explore → prometheus → hephaestus → testing-expert → code-reviewer → librarian
Use for: Complete feature from scratch with full quality gates.
→ hephaestus (feature A)
root → hephaestus (feature B) → code-reviewer
→ librarian (docs)
Use for: Independent features that can be built in parallel.
implement → code-reviewer → fix issues → re-review
Use for: Ensuring high quality on critical code.
explore → atlas → prometheus → sisyphus → testing-expert
Use for: Large projects on unfamiliar codebases.
Agents use the model configured in settings.json modelProviders. The current
setup assigns all agents to the qwen-oauth provider with model selection handled
by Qwen Code's model resolution.
Include these phrases in your commands to encourage proper delegation: