error-recovery

Error recovery patterns for squad agents enable adaptive behavior during API failures by applying retry with backoff, fallbacks, and graceful degradation to preserve overall task value instead of failing silently.

You implement deterministic retry and fallback strategies by activating standardized recovery templates in agent tooling, which automatically guide retries, alternate methods, and documented recovery paths during tool malfunctions.

Yes, you can apply structured fault tolerance patterns directly to build and test errors, utilizing diagnose-and-fix workflows to identify root causes and apply remedies before triggering clear escalation paths.

Graceful degradation handles non-critical step failures by allowing squad agents to bypass or minimize those errors, preserving the overall task value and ensuring the primary workflow continues despite minor disruptions.

You should use escalation paths when diagnose-and-fix workflows cannot resolve the root cause, handing off unresolved issues with full context to a human operator or a coordinator agent.