dispatching-parallel-agents

Parallel agent dispatching runs separate AI agents concurrently on independent test failures, eliminating sequential debugging delays. Each agent focuses on one problem domain, allowing you to investigate and resolve multiple unrelated failures simultaneously instead of one at a time.

Use parallel agents when you have 3+ independent failures across different test files or subsystems with no shared state. Each failure must be analyzable and fixable independently. If failures share dependencies or require coordinated fixes, sequential debugging is more appropriate.

Failures are independent if they occur in different test files or subsystems and don't share state or dependencies. The Skill provides a flowchart to evaluate independence. Isolated root causes that don't require coordinated code changes indicate failures suitable for parallel agent dispatch.

Each agent should produce concise summaries suitable for integration. Use event-based progress signals instead of arbitrary timeouts, and ensure outputs are focused summaries of the investigated issue and fix applied, ready to merge back into the codebase.

Parallel dispatching works best when failures are truly isolated across problem domains with strict change isolation. Success depends on having clearly separable test files or subsystems and the ability to assign one focused agent per independent failure without cross-contamination.