code-review

Review diffs and pull requests for bugs, security issues, and merge blockers.

1|Updated Apr 24, 2026
One-click install
npx skills add https://github.com/kreek/consult --skill code-review-kreek
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/kreek/consult/tree/main/plugin/skills/code-review
Command: npx skills add https://github.com/kreek/consult --skill code-review-kreek

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Agent-generated and human-written code changes often ship with hidden defects, unproven claims, and security gaps that surface only in production. This Skill performs independent, findings-first review of diffs and PRs so correctness, safety, and data-loss risks are caught before merge. ## Core Features & Use Cases - Severity-ranked findings: Every issue is reported with a file/line anchor, impact, fix direction, and evidence, ordered Critical through Low, with summaries after findings. - Independent fresh-context review: Runs in a subagent, review agent, or separate session given only the diff, stated intent, and repo constraints, avoiding author-rationale bias. - Language-specific reviewer guides: Loads per-language references for Rust, Python, TypeScript, Ruby, Java, Kotlin, C#, F#, Bash, and SQL, plus security and proof-evidence passes on every review. - Use Case: After an agent implements a feature and tests pass, run this Skill on the branch diff to catch a missing authorization check, an unproven performance claim, and a speculative abstraction before requesting human merge approval. ## Quick Start Use the code-review skill to review my current branch diff for blocking correctness, security, and maintainability issues.

Frequently Asked Questions about code-review

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

FAQPage Schema
How do I review a GitHub pull request with an AI agent?▼

Point the code-review skill at the PR through your host's GitHub surface. It pre-flights intent, CI status, and changed surface, then produces severity-ranked findings with file anchors. Writes like comments or reviews require explicit user permission.

What makes a code review independent instead of same-context?▼

Independent review runs in a fresh context such as a subagent, review agent, or separate session, receiving only the diff, stated intent, and repo constraints. Withholding implementation rationale prevents reviewer bias; same-context review is a weaker labelled fallback.

Which programming languages does this code review support?▼

It ships dedicated reviewer references for Rust, Python, TypeScript, Ruby, Java, Kotlin, C#, F#, Bash, and SQL. Each guide covers tooling gates, high-signal checks, and anti-patterns specific to that language and its ecosystem.

When should I not use a general code review skill?▼

Skip it for commit grouping or history repair, which belongs to git-workflow, and for narrow domain-only reviews where a dedicated security, database, API, or UI-design skill is sufficient. It targets defect discovery, not proof generation.

Why does the review report unproven claims instead of passing them?▼

A claim the reviewer cannot verify, such as red or absent CI or missing test evidence, is reported as unproven rather than accepted. This keeps passing tests from being treated as proof of safety, data integrity, or behavior correctness.