--- description: Classify coder-loop failures and choose the next recovery strategy agent: eoc_orchestrator subtask: true --- # Failure Strategy Command Use this after a failing coder-loop round to decide whether to keep fixing, rebuild context, repair the environment, review, or roll back. ## Core Usage ```bash # Inspect the latest coder-loop run node bin/eoc-script.js failure-strategy report # Inspect a specific run in machine-readable form node bin/eoc-script.js failure-strategy report --run-id --json ``` ## What It Decides - whether the current failures are local and worth another fix round - whether the context is stale and should be rebuilt - whether the failure is environmental or dependency-related - whether a linked snapshot makes rollback safer than continuing - whether validation is green and the task should move to `/review-gate` ## Signals It Uses - failing check kinds and failure categories - number of touched files and blast radius - repeated identical failure fingerprints across rounds - regression spikes between rounds - environment and dependency error patterns - linked implementation plan and safety snapshot availability - benchmark-aware risk feedback for the current runtime/framework/task family