skills.rest
Cain-Ish
Cain-IshCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

code-review-deep

Perform multi-pass code reviews on GitHub pull requests using local git diffs.

Updated Apr 24, 2026
Search Tags:#diff analysis#architecture#code review#pull request#github#regression detection#second-brain
One-click install
npx skills add https://github.com/Cain-Ish/claude-code-plugin --skill code-review-deep
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-review-deep
Source: https://github.com/Cain-Ish/claude-code-plugin/tree/main/skills/code-review-deep
Command: npx skills add https://github.com/Cain-Ish/claude-code-plugin --skill code-review-deep

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It eliminates missed bugs, regressions, and architectural risks in pull requests by producing a deep, structured, multi-pass review from a local git checkout and PR metadata.

Core Features & Use Cases

  • Multi-pass decomposition: breaks a change into logical review units (tests, modules, feature slices, and infra) and reviews each with the right model selection strategy.
  • Architecture risk scanning: adds an advisory architectural pass focused on the highest-risk units rather than trying to reason about everything at once.
  • Regression-aware history: performs a regression/history pass and scores findings to separate confirmed issues from lower-confidence candidates, reducing both false alarms and blind spots.
  • Second-brain conventions: consults local conventions and prior review dismissals to align with project patterns and remembered false positives.
  • PR posting or terminal output: can print to the terminal by default, or post results as a PR comment with consistent formatting and links to exact lines.

Quick Start

Run code-review-deep on your current branch by issuing: review the current branch against its base and print the deep review to the terminal.

Frequently Asked Questions about code-review-deep

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

FAQPage Schema
How do I perform a deep code review on a GitHub pull request using a local git checkout?▼

A deep code review on a GitHub pull request uses your local git checkout to decompose diffs into logical units, run architecture risk triage, check regression history, and output line-linked findings. It scores issues to suppress false positives.

What is multi-pass decomposition in pull request diff analysis?▼

Multi-pass decomposition in pull request diff analysis breaks changes into logical review units like tests, modules, and feature slices. Each unit is reviewed with the appropriate model selection strategy to isolate architectural risks and regressions.

Can I post code review findings as a GitHub PR comment with exact line references?▼

Yes, you can post code review findings as a GitHub PR comment. The review output includes consistent formatting with full SHA line references that link directly to the exact lines changed in the pull request diff.

How does regression detection work during a code review?▼

Regression detection during a code review performs a history pass to identify past issues and scores findings to separate confirmed problems from lower-confidence candidates. This scoring reduces false alarms while preventing blind spots.

Does the code review tool align with local project conventions and past dismissals?▼

Yes, the code review tool consults a second-brain system to retrieve local project conventions and prior review dismissals. This aligns the pull request analysis with established project patterns and remembered false positives.

What is the best way to analyze architectural risks in a large pull request?▼

The best way to analyze architectural risks in a large pull request is an advisory architectural pass focused on the highest-risk logical units. This targeted approach prevents reasoning overload and highlights critical structural issues.

Related Skills

Explore More Software Engineering Skills→
OlshanskOlshansk
8

cmd-pr-review-prepare

Generate context-rich git diffs and identify code improvement areas.

Community
Intermediate
wyattowalshwyattowalsh
4

honest-review

Analyze code correctness, design, and efficiency with confidence-scored findings.

Community
Advanced
yongwoonyongwoon
8

ywc-review-learnings

Manage project-specific code-review preferences in a version-controlled Markdown file.

Community
Advanced
davduncdavdunc

Evals

Run regression and capability evaluations for multi-turn AI agent workflows.

Community
Advanced