| name | parallel-launch |
| description | Use when the request has multiple independent sub-tasks, cross-domain research, or parallelizable file or module work. |
Parallel Launch
Decompose the given task into independent agent groups and execute them in broad parallel.
Process
-
Analyze the task and identify independent concerns that can run concurrently.
- Each concern must be self-contained: no shared mutable state, no ordering dependency.
- If concerns have dependencies, sequence the dependent batch after the independent batch completes.
- Consult delegation scenarios for parallelism decisions.
-
Design agent groups for each independent concern:
- Assign a clear, scoped objective (one concern per agent).
- Select the appropriate tailored agent type (Explore, Plan, or a domain specialist) — prefer a task-specific agent over a general one.
- Define expected output format so results can be composed.
-
Launch all independent agents in a single tool call.
-
Compose results once all agents complete:
- Merge non-conflicting outputs directly.
- For conflicting or overlapping results, reconcile and present trade-offs to the user.
- If any agent failed or returned incomplete results, report the gap and propose a targeted follow-up.
-
Review composed output by dispatching a review agent to verify:
- Completeness: All original concerns addressed, no gaps.
- Consistency: No contradictions between agent outputs.
- Accuracy: Claims are substantiated, sources checked, no hallucinated findings.
- Scope: Nothing extra built beyond what was asked.
- For implementation work, additionally verify spec compliance and code quality.
-
Report to user only after review passes.
Constraints
- Agents per batch: match the number of truly independent concerns (avoid artificial splitting).
- Each agent prompt must include full context; agents do not share memory.
- If the task has fewer than 2 independent concerns, execute directly instead of launching agents.
