---

name: dispatching-parallel-agents description: > Use this skill when you have multiple independent tasks or failures that can be investigated in parallel without shared state.

Dispatching Parallel Agents

Intent

Reduce cycle time by delegating independent investigations or tasks to parallel agents while preventing scope collisions.

---

When to Use

• Multiple independent failures or tasks exist.
• The tasks do not share state or files.
• The output can be integrated without ordering dependencies.

---

Precondition Failure Signal

• Parallel agents modify the same files or components.
• Tasks are not clearly scoped.
• Integration requires rework because outputs conflict.

---

Postcondition Success Signal

• Each agent has a focused scope and clear constraints.
• Outputs are reviewed and integrated without conflicts.
• Full verification passes after integration.

---

Process

1. Partition: Group work into independent domains.
2. Scope: Define precise tasks, constraints, and expected outputs.
3. Dispatch: Assign one agent per domain.
4. Review: Validate each result for scope and correctness.
5. Integrate: Merge outputs and run full verification.

---

Example Test / Validation

• Three independent test failures are investigated in parallel and the suite passes after integration.

---

Common Red Flags / Guardrail Violations

• "Fix everything" prompts with no constraints.
• Parallel changes to shared files.
• Skipping integration verification.

---

Recommended Review Personas

• Platform Engineer - validates integration and verification.
• Tech Lead - validates scope boundaries.

---

Skill Priority

P3 - Delivery & Flow

---

Conflict Resolution Rules

• If tasks are related, prefer systematic-debugging over parallelisation.
• Safety and correctness skills override speed.

---

Conceptual Dependencies

• systematic-debugging
• verification-and-handover

---

Classification

Delivery
Operational

---

Notes

Parallelism is a tool, not a default. Use only for truly independent work.