Dev builds triage skill to tame GPT-6.2's overly paranoid code review

PawelHuryn · x · 2026-11-21

Pawel Huryn shares that GPT-6.2 is an excellent code reviewer but flags nearly everything as Medium or High severity. His fix: an independent-review skill that interrogates each finding — who hits it and how often, the harm in one sentence, and whether it crosses a red line (other users' data, lost work, unapproved irreversible writes).

On one build, nine review rounds produced noise; round ten with triage yielded 11 findings, none above Low, closing the stage with zero fixes needed.

The skill is part of his open-source product-engineering repo: 45 agent skills covering decide → build (Claude Code/Codex) → second-model review → release. MIT licensed, 687 stars.

Related event: Developer Builds Independent Review Skill to Tame Overcautious GPT-6.2 Code Reviews(3 posts)→

Original post →

More from coding & agent

coding & agent channel →