root-cause

Root cause analysis on code review findings uses a compressed 5-why technique to challenge proposed fixes, classifying whether they address underlying causes or merely patch symptoms. It filters for high and medium confidence scored findings to prevent wasted effort on false positives.

Batching code review findings groups them by file and containing function, pre-reads functions with surrounding context, and pre-greps caller chains. This source-aware grouping provides necessary code context before spawning parallel root-cause analysis agents per batch.

Compressed 5-why root-cause analysis in software engineering is a technique that challenges proposed fixes to determine if they address the underlying root cause. It helps teams avoid repeated regressions and misguided fixes by applying structured questioning to scored findings.

Yes, root-cause analysis requires read access to a SCORED_FINDINGS.md file containing previously scored findings. It automatically selects high and medium confidence findings from this file to focus analysis effort on likely real issues rather than false positives.

Merging parallel root-cause analysis results validates batch writes and combines them into a single ROOT_CAUSE_ANALYSIS.md file. The system enforces a strict structured output format per finding during parallel agent execution before merging the validated batch outputs.

Automated 5-why analysis should not be used when you lack read access to repository source files or a SCORED_FINDINGS.md file. It also requires the ability to write batch files and a final ROOT_CAUSE_ANALYSIS.md, making it unsuitable for restricted environments.