coder-rules-debug-requirements

Investigating functional bugs, crashes, regressions, or production incidents Need structured debug flow for coding agent execution Need explicit completion criteria for bug fixes </when_to_use> Symptom statement and impact Expected vs actual behavior Minimal reproducible example (steps/data) Environment fingerprint (version, config, flags) Evidence pack (errors, logs, traces, timestamps) </input_requirements> Triage impact and confirm current reproducibility Build minimal reproducible case Localize failing layer and narrow search space State testable hypothesis and instrumentation plan Apply minimal root-cause fix Verify fix against acceptance criteria Add regression test/eval that fails before and passes after Record residual risks and follow-up actions </debug_workflow> No reproducible signal means no claim of confirmed fix Bugfix is incomplete without regression protection Fix must target root cause, not only symptom masking Verification must include evidence from commands or telemetry </quality_rules> Repro rate before and after fix Error rate or failure count change Latency/perf guardrail impact when relevant Regression suite result for affected path </required_signals> Do not debug by guesswork without hypothesis Do not close issue without reproducible verification Do not skip regression test because fix "looks obvious" Do not merge broad refactor as hidden hotfix </do_not> Problem statement with expected vs actual Root cause summary and impacted components Fix summary with evidence of verification Regression artifact added and location </output_requirements> Stack Overflow MRE Guide MIT Debugging Workflow OpenAI Evaluation Best Practices OpenAI Evals Guide Anthropic: Demystifying Evals for AI Agents Google SRE Postmortem Culture OpenTelemetry Context Propagation