dispatching-parallel-agents

Dispatch parallel agents to investigate independent failures concurrently. Create one agent per test file, subsystem, or bug domain with narrowly scoped prompts and required context to prevent shared context contamination and enable simultaneous investigation.

Parallel agent orchestration is a pattern to dispatch focused, self-contained agents so independent problems can be explored and fixed concurrently. It enforces explicit scopes, self-contained contexts, and post-return review steps to detect conflicts and integrate changes safely.

Dispatch focused agents by creating one agent per independent domain with a narrowly scoped prompt and required context. Run investigations concurrently, review summaries, check for conflicts, run the full test suite, and integrate non-conflicting fixes.

Yes, you can use parallel agents for concurrent subsystem debugging when domains do not share state or sequential dependencies. Apply this pattern when separate test files, subsystems, or bug domains can be debugged concurrently without interference.

Avoid parallel agents when independent failures share state or have sequential dependencies. If bug domains require shared context or sequential investigation, sequential debugging is more appropriate to prevent context contamination and integration conflicts.

Integrate non-conflicting fixes by reviewing agent summaries post-return, checking for conflicts, running the full test suite, and merging changes. This post-return review step detects conflicts and ensures integrated fixes do not break dependent subsystems.