error-recovery

Handle transient API failures in agent workflows by applying retry with backoff patterns, which systematically re-attempt failed requests to improve reliability without manual intervention.

The best way to maintain partial functionality when a component fails is to apply fallback and graceful degradation patterns, allowing agent workflows to continue operating in a reduced capacity instead of crashing completely.

Systematically diagnose and fix build or test errors by applying diagnose-and-fix guidance patterns, which provide structured steps to identify root causes and remediate operational faults across squad agents.

Use escalation with context for operational faults when automated recovery patterns fail, ensuring unresolved issues are handed off to humans or specialists with the necessary logging and criteria for remediation.

Yes, this error-handling approach works across multiple squad agents by standardizing recovery patterns for retry, fallback, and graceful degradation to ensure operational faults are addressed consistently across the entire workflow.

Automatic recovery patterns for agent workflows are limited by fault type; while they standardize handling for transient failures and build errors, unresolved issues still require manual escalation with context to a human specialist.