crit

Review local diffs and pull requests for concrete defects and risks.

1|1|Updated Apr 24, 2026

Search Tags:#code review #security review #risk assessment #adversarial testing #correctness #git diff #PR verification

Ready to run

npx skills add https://github.com/luan/agents --skill crit-luan

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: crit
Source: https://github.com/luan/agents/tree/main/skills/crit
Command: npx skills add https://github.com/luan/agents --skill crit-luan

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Crit helps you catch genuine defects, integration issues, and security or correctness risks in local diffs, branch changes, or PRs—without inventing new work or reviewing unrelated code.

Core Features & Use Cases

Adversarial diff-based review focused only on issues caused or exposed by the change.
Deterministic reviewer dispatch with a predefined set of focus blocks and structured severity output.
Guardrails for large diffs and bugfix-only changes, including truncation handling and a FAIL verdict when a “bugfix” touches only test code.

Quick Start

Ask the AI to run the crit skill on your PR diff and return a verified severity table with only concrete failure modes and trigger conditions.

Frequently Asked Questions about crit

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

FAQPage Schema

How do I perform an adversarial code review on a PR diff?▼

To review a pull request diff, request an adversarial assessment that identifies concrete defects and risk triggers. It resolves BASE context, runs focused sub-reviewers on the raw diff, and outputs a verified severity table with a PASS, CHANGES_REQUESTED, or FAIL verdict.

What is adversarial risk assessment for git diff changes?▼

Adversarial risk assessment evaluates git diff changes strictly for issues caused or exposed by the modification. It avoids inventing unrelated work by applying deterministic reviewer dispatch to find correctness, security, and operational risks across varying diff shapes.

How do you handle code review for large branch changes?▼

Code review for large branch changes handles scale by truncating large file reads with mandatory verification. It ensures comprehensive risk assessment by applying guardrails and dispatching focused sub-reviewers directly on the raw diff to manage operational load.

How do I verify a bugfix PR only touches test code?▼

To verify a bugfix PR, adversarial review applies specific guardrails that trigger a FAIL verdict if a change labeled as a bugfix touches only test code. This ensures the actual defect in the source code is addressed.

What security review checks should I run on local diffs?▼

Security review on local diffs should identify concrete defects and risk triggers without reviewing unrelated code. By dispatching focused sub-reviewers on the raw diff, it targets security, correctness, and operational vulnerabilities exposed by the change.

What is the best way to assess correctness in a pull request?▼

The best way to assess correctness is running an adversarial review that resolves BASE context and verifies failure modes. It evaluates integration and maintainability risks specifically caused by the PR changes to produce a verified severity table.