| name | cc-orchestrator |
| description | General-purpose development task orchestrator. Breaks complex development tasks down into specialized agent roles and coordinates their collaboration. Use it when you need to 'form a team', 'decompose tasks', 'multi-agent collaboration', 'team-based development', 'parallel development', or 'team collaboration'. Follow-up tasks supported: 'modify the team', 'adjust the division of work', 're-run', 'update the plan', 're-run only a specific phase'. For simple tasks or single-file changes, just do them directly; only enable this when complexity genuinely requires multi-role collaboration. |
| metadata | {"version":"1.0.0"} |
Team Orchestrator — general-purpose development task orchestrator
Break a complex development task down into specialized agent roles, pick a suitable architecture pattern, and coordinate a team of agents to complete the task together.
Core principles:
- An agent team is the default execution mode — prefer TeamCreate whenever there are 2 or more members.
- Role definitions are separate from orchestration — agents are defined in
agents/, the orchestration logic lives in this skill.
- Intermediate artifacts all go in
_workspace/ — naming convention {phase}_{agent}_{artifact}.{ext}.
When to enable
- The user explicitly asks for "form a team", "multi-agent collaboration", "decompose the task", or "push forward in parallel".
- The task contains at least 2 subproblems that need different areas of expertise.
- The task needs intermediate design, review, or test artifacts explicitly preserved for later traceability.
Do NOT enable this orchestrator in the following cases — just do the work directly:
- Simple Q&A, small single-file changes, local bug fixes.
- Tasks that can be finished directly in the current session without a division of roles.
- You only need a single expert perspective rather than multi-role collaboration.
Execution modes
| Mode | Applicable scenario | Tools |
|---|
| Agent team (default) | 2+ agents need to collaborate and cross-check | TeamCreate + SendMessage + TaskCreate |
| Sub-agent | A single task, only the result needs to be returned | Agent tool + run_in_background |
| Hybrid | Phases have different characteristics | Switch modes per phase |
See references/pattern-selector.md for the detailed mode-selection guide.
Workflow
Phase 0: Context confirmation
- Check whether the current project's
_workspace/ directory exists.
- Determine the execution mode:
_workspace/ does not exist → initial run, proceed to Phase 1.
_workspace/ exists + user requests a partial change → partial re-run (only re-run the relevant agents).
_workspace/ exists + brand-new input → new run (back up the old _workspace/ as _workspace_{YYYYMMDD_HHMMSS}/).
- Identify the project's tech stack (check pyproject.toml / package.json / Cargo.toml / go.mod, etc.).
Phase 1: Task analysis
- Understand the user's task goals and constraints.
- Determine the nature of the task (new feature / research / refactor / fix / docs / review).
- Estimate the task size:
- Small (1-2 concerns) → no team needed, sub-agent or do it directly.
- Medium (3-5 concerns) → 2-3 person team.
- Large (6+ concerns) → 3-5 person team.
- Pick an architecture pattern per
references/pattern-selector.md.
- Pick the agent role combination per
references/agent-catalog.md.
- Decide whether explicit confirmation is needed:
- When requirements are unclear, approaches diverge significantly, execution cost is high, or impact could be large → confirm with the user first.
- When the user has already explicitly asked for team execution and the approach is clear with controllable risk → proceed directly to Phase 2.
Phase 2: Team formation
- Form the team based on the Phase 1 analysis:
TeamCreate — members reference the role definitions in the agents/ directory.
TaskCreate to register subtasks:
- Assign 3-6 tasks per agent.
- Use
depends_on to declare task dependencies.
- Create the
_workspace/ directory to store intermediate artifacts.
Phase 3: Execution
- Members claim and execute tasks via the shared task list.
- Members coordinate in real time via
SendMessage:
- Share findings, discuss conflicts, fill in gaps.
- They can communicate directly without going through the Leader.
- The Leader monitors progress via
TaskGet and steps in to coordinate when necessary.
- Artifacts are saved to
_workspace/{phase}_{agent}_{artifact}.{ext}.
Phase 4: Integration and verification
- Collect every member's artifacts via
Read.
- Check consistency:
- Do both sides of an interface definition match?
- Are there contradictions between artifacts?
- Produce the final artifact.
- Brief quality verification.
Phase 5: Cleanup and reporting
- Report completion status and the location of key artifacts to the user.
TeamDelete to clean up the team.
- Keep
_workspace/ for auditing and follow-up tasks.
Common team templates
Feature development
architect(1) → developer(1-2) → reviewer(1) + tester(1)
Pattern: pipeline + generate-verify
Technical research
researcher(2-4)
Pattern: fan-out/fan-in
Code review
reviewer(2-3) (each focusing on security / performance / architecture)
Pattern: fan-out/fan-in
Large-scale refactoring
project-lead(1) + developer(2-3)
Pattern: supervisor
Full-stack development
architect(1) + developer-frontend(1) + developer-backend(1) + tester(1)
Pattern: hierarchical delegation
Documentation improvement
tech-writer(1) + researcher(1)
Pattern: fan-out/fan-in
Error handling
| Situation | Strategy |
|---|
| A single member fails | Leader confirms via SendMessage → restart or create a replacement |
| More than half the members fail | Notify the user to confirm whether to continue |
| Timeout | Use the partial results already collected |
| Data conflict between members | Annotate the sources side by side; don't delete either one |
| Delayed task status | Leader confirms via TaskGet, then manually TaskUpdate |
Data-passing conventions
| Strategy | Method | Applicable scenario |
|---|
| Message | SendMessage | Real-time coordination, lightweight feedback |
| Task | TaskCreate/TaskUpdate | Progress tracking, dependency management |
| File | Write/Read _workspace/ | Structured artifacts (>100 lines) |
| Return value | Agent tool return | Sub-agent result collection |
Conflict-resolution protocol
- Data conflict → don't delete either side; annotate the sources and list them side by side.
- Design disagreement → the architect has the final say, but must explain the reasoning.
- Implementation diverges from design → the developer reports to the architect, who makes the call.
Test scenarios
Normal flow: feature development
- User requests "add user authentication to the project".
- Phase 1 analysis → medium-sized feature development → pipeline pattern.
- Phase 2 forms an architect + developer + reviewer team.
- Phase 3 executes in order: design → implement → review.
- Phase 4 integration.
- Output: design doc + code + review report under
_workspace/.
Error flow: development blocked
- During Phase 3 the developer hits an unclear interface definition.
- The developer SendMessages the architect to request clarification.
- The architect fills in the interface definition and SendMessages a reply.
- The developer continues the implementation.