Score candidate patches against a rubric using test outcomes and diff analysis.
npx skills add https://github.com/paulingham/.claude --skill patch-critique
Please help me install this Agent Skill. Skill: patch-critique Source: https://github.com/paulingham/.claude/tree/main/skills/patch-critique Command: npx skills add https://github.com/paulingham/.claude --skill patch-critique
Evaluates candidate patches at the Final Gate by analyzing test results and the patch diff to decide PATCH_APPROVED or PATCH_REJECTED.
Provide the candidate patch diff and the latest test results to the patch-critic and run the Final Gate evaluation.
High-intent search queries and answers about installing and using this skill.
Automating code review for patch diffs involves scoring candidate patches at the Final Gate by analyzing test outcomes and diffs to decide PATCH_APPROVED or PATCH_REJECTED. It uses a patch-critic agent pairing to provide multi-dimensional rubric judgments for automated pipelines.
A Final Gate evaluation is a decision point that scores candidate patches by analyzing test outcomes and the patch diff to output PATCH_APPROVED or PATCH_REJECTED. It operates with a Sonnet executor and Opus advisor agent pairing to provide multi-dimensional rubric judgments.
Evaluating code changes using a multi-dimensional rubric requires spawning a patch-critic agent to score the patch against four dimensions. It integrates test results, candidate diff, intake spec, and verification evidence to drive gating decisions in automated pipelines.
Yes, this Skill operates with a patch-critic agent pairing that uses a Sonnet executor and Opus advisor. This pairing scores candidate patches against four dimensions by analyzing test outcomes and patch diffs to support decision-making in automated pipelines.
To score candidate patches, you need to provide the candidate patch diff and the latest test results to the patch-critic. The evaluation also requires access to the intake task description and verification evidence to drive gating decisions.
Automated patch gating is not suitable when you lack the required inputs, such as candidate diffs, latest test results, intake task descriptions, or verification evidence. The patch-critic requires all these data sources to perform its multi-dimensional rubric judgment effectively.
kevinnguyen271090Enforce code-review governance with verification gates before completion claims.
ramontramontiniExecute downstream stories from ready_to_pull to done with enforced gates.
bennybennisonRun sequential build, type, lint, test, security, and diff checks.
ChakshuGautamEvaluate agent outputs against multi-dimensional rubrics to produce scores and pass/fail judgments.
