system-parallel-agents

To fix independent test failures concurrently, dispatch parallel agents where each agent receives self-contained prompts with explicit scope and constraints to investigate isolated subsystems. This pattern allows unrelated debugging domains to be analyzed simultaneously without cross-contaminating state.

The best way to structure prompts for parallel debugging agents is to provide minimal context, explicit constraints, and clear expected outputs for each independent task. Each self-contained prompt must define a distinct scope so agents can run concurrently without sharing state.

To integrate fixes after running parallel agents, review the concise change summaries from each agent, check for conflicts between proposed fixes, run the full test suite, and integrate the verified changes into the main codebase.

Use parallel agents for debugging when multiple unrelated test failures, subsystems, or debugging domains can be analyzed without shared state. If independent problems can run concurrently without cross-contamination, parallel execution resolves them faster than sequential investigation.

Yes, you can use parallel agents to investigate failures across separate test files during refactoring. Dispatch one agent per failing test file so each agent concurrently identifies root causes, proposes fixes, and returns concise change summaries for integration.

Parallel agent debugging requires independent problems with no shared state; if failures are interdependent or require cross-domain context, isolated agents cannot coordinate effectively. Each agent needs fully self-contained prompts, making complex coupled bugs unsuitable for this approach.