skills.rest
mimukit
mimukitCOMMUNITY
Explore all skills by this author →
Website𝕏@mimukitGitHub

reviewkit

Reviews AI-generated code changes in four ordered passes against the working tree or branch diff.

1|Updated Jul 12, 2026
Search Tags:#static analysis#code review#pull request#git diff#ai-generated code#severity ranking#convention-fit
One-click install
npx skills add https://github.com/mimukit/skills --skill reviewkit-mimukit
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: reviewkit
Source: https://github.com/mimukit/skills/tree/main/skills/reviewkit
Command: npx skills add https://github.com/mimukit/skills --skill reviewkit-mimukit

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI agents produce code that is correct in a vacuum but wrong for the repo, padded with plausible-looking cruft, or quietly incomplete. Generic bug-focused reviews miss these failure modes, and self-review by the agent that wrote the code tends to rubber-stamp it. ## Core Features & Use Cases - Four ordered review passes: convention-fit against repo idioms, agent-slop signatures (over-engineering, dead code, hallucinated APIs, scope creep), requirement-completeness against the stated intent, and classic correctness including test-quality checks. - Evidence-backed findings: every finding is ranked by severity (Blocker, Should-fix, Nit), quoted from the diff, and tied to what it violates, with a mechanical verdict and a coverage note naming what was not verified. - Fresh-eyes delegation: hands the diff to a subagent with no memory of the implementation session, or clearly labels the result a self-review when no subagent is available. - Use Case: After an agent finishes a feature branch, ask for a review before opening a PR; the skill resolves the base ref via gitkit, reads the full diff including untracked files, and returns a ranked report you can optionally save to docs/reviews/ for the PR description. ## Quick Start Ask the agent to review the changes it just made on this branch before committing, and have it report findings ranked by severity.

Frequently Asked Questions about reviewkit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review code written by an AI agent before committing?▼

Run a structured review of the working tree or branch diff that checks convention-fit, agent-slop patterns, requirement-completeness, and correctness in order. Ground the review in the actual git diff, including untracked files, and require quoted evidence for every finding.

What are common signs of AI-generated code slop?▼

Typical signatures include over-engineering for single callers, dead or unreachable code, hallucinated API calls, redundant comments restating the code, scope creep beyond the request, and fake robustness like error-swallowing try/catch blocks or tautological tests.

Should the same AI agent review code it wrote itself?▼

No, self-review tends to rubber-stamp decisions the agent already rationalized. Delegate the review to a fresh subagent with no memory of the implementation session; if none is available, label the output explicitly as a self-review so readers can weight it.

Does this code review skill edit or fix the code it reviews?▼

No, it is read-only by contract. It runs git and search commands plus optionally the repo's test command, never edits source, and its only write is an optional review report saved to docs/reviews/ when the user asks.

How does the review handle untracked files in git?▼

git diff does not show untracked files, so the review lists them with git status --porcelain and reads every untracked source file in full. Only generated files like lockfiles and build output are skipped after a spot-check.

Related Skills

Explore More Software Engineering Skills→
NauelGNauelG
28

review-change

Reviews a git diff against coding standards and the originating spec using parallel sub-agents.

Community
Intermediate
Strategic-AutomationStrategic-Automation
6

requesting-code-review

Automate pre-commit verification with security scans, quality gates, and reviewer subagent.

Official
Advanced
arbazkhan971arbazkhan971
26

lint

Configure linters, formatters, pre-commit hooks, and CI enforcement across repositories.

Community
Advanced
withpiperwithpiper
6

release-changelog

Parse git commits between refs to draft a structured CHANGELOG entry.

Official
Advanced