iterate

Autonomous multi-cycle debugging resolves integration failures by continuously diagnosing root causes, applying minimal fixes, and verifying outcomes with concrete checks until acceptance criteria are satisfied.

Root cause analysis verification uses repeated testing cycles that collect evidence through API calls, DB queries, and build commands rather than assumptions. It applies guarded stopping conditions to prevent infinite loops during end-to-end testing.

Iterative refinement runs structured diagnose, fix, test, and evaluate phases until acceptance criteria are satisfied. It applies minimal changes and validates outcomes through evidence collection with optional subagent parallelization for concurrent tests.

Subagents can parallelize broad codebase research and concurrent test groups during debugging iterations. The core reasoning and decisions remain preserved while subagents assist with gathering evidence for complex multi-step diagnosis.

Autonomous iteration loops should not be used when guarded stopping conditions are absent, as infinite loops may occur without clear acceptance criteria. It requires structured phase execution and evidence collection to function properly.