What problem does it solve? Before creating a pull request, a single review-and-fix pass often leaves residual issues or produces shallow local patches. This Skill automates an iterative loop of deep review and root-cause correction on your local branch diff until no findings remain, so the PR is genuinely ready. ## Core Features & Use Cases - Iterative Review-Fix Loop: Runs pr-deep-review style reviews each round via a dedicated subagent or codex review, then applies upstream-fix root-cause analysis across all findings until convergence. - Cluster-Based Diagnosis: Groups findings by shared design root causes (responsibility placement, normalization gaps, ambiguous contracts) and fixes each cluster as one coherent change rather than scattered patches. - Convergence Gating: Records each round's finding count via a hook script, supports semantic mid-loop convergence with user acceptance, and enforces round limits with explicit user confirmation. - Use Case: Before running gh pr create, ask the AI to loop review and fixes on your branch; it reviews the diff against origin/main, fixes root causes round by round, and reports a per-round summary with a readiness verdict. ## Quick Start Ask the AI to review and fix the current branch diff repeatedly until no findings remain before creating the pull request.
