Reviews pull request diffs against an ideal bar using adversarial verification workflows.
npx skills add https://github.com/srid/emanote --skill perfection-review
Please help me install this Agent Skill. Skill: perfection-review Source: https://github.com/srid/emanote/tree/main/.agents/skills/perfection-review Command: npx skills add https://github.com/srid/emanote --skill perfection-review
Standard code reviews settle for "correct enough" and miss defects that resurface one seam over after each fix. This Skill holds a change to an ideal bar — where the defect can no longer be expressed — and tracks it across review rounds until it has nowhere left to relocate.
Ask the AI to run a perfection review on PR 42 against the main branch and report the residual defect surfaces with structural fixes.
High-intent search queries and answers about installing and using this skill.
Invoke this Skill with a PR number to hold the change to an ideal bar where the defect can no longer be expressed. It grounds every claim in the diff, hunts where the defect relocates, and reports residual surfaces with structural fixes.
Adversarial review fans out fresh-context verifier agents, one per claim, plus an isolated adversary whose only job is to express the defect anyway. Each agent cites the diff, and results are synthesized with a refuted-if-uncertain default.
Use it only when explicitly seeking an ideal-bar review, such as making a defect impossible to express or tracking where a defect relocates across rounds. It assumes eternal time and no ship pressure, so it is not suited for routine reviews.
Yes, but only when explicitly requested with the --post flag. By default the review reports findings without posting, leading with credit for what is closed before listing ranked residual surfaces.
A fix that satisfies the literal ask often lets the same defect shape resurface one seam over. This Skill names the defect's shape and tracks it across rounds until it has nowhere left to relocate.
FrogAiGenerate PR descriptions from live templates, commits, diffs, and linked tickets.
code-review-benchmarkAutomate local diff reviews with CodeRabbit until zero findings remain.
mthinesAnalyze test suites to verify imports and exercise of production functions.
antonyfmunozRun adversarial code review on changed source files and verify with py_compile.
