systematic-debugging

To find the root cause of repeated task failures in multi-agent workflows, apply a systematic four-phase debugging methodology: Observe, Narrow, Hypothesize, and Fix. This structured approach replaces guesswork with evidence collection, cross-agent checks, and escalation paths to pinpoint failures.

The best way to debug distributed agents that fail repeatedly is to trigger a systematic debugging workflow after three worker failures. This initiates phase-driven diagnosis, applying cross-agent collaboration checks like git blame and task dependency analysis to isolate the fault.

The four-phase debugging process for distributed agents involves Observe to collect evidence, Narrow to isolate variables, Hypothesize to identify the root cause, and Fix to apply the solution. Each phase generates explicit outputs and defensive recommendations.

To handle uncertain test-failure analyses across distributed agents, use a systematic debugging skill that provides structured escalation paths and cross-agent guidance. It helps verify findings when a mend-fixer cannot reproduce the issue, ensuring accurate root-cause identification.

If ward checks fail after a fix in a multi-agent system, the initial repair likely missed the root cause. A systematic debugging approach addresses this by returning to evidence collection and hypothesis formation to ensure the underlying issue is fully resolved.