agent-review

Reviews pull requests against project standards and posts structured feedback on GitHub.

1|Updated Feb 8, 2026
One-click install
npx skills add https://github.com/blamechris/skill-templates --skill agent-review-blamechris
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-review
Source: https://github.com/blamechris/skill-templates/tree/main/.claude/skills/agent-review
Command: npx skills add https://github.com/blamechris/skill-templates --skill agent-review-blamechris

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Code reviews often miss project-specific conventions, leave deferred suggestions untracked, and produce inconsistent feedback. This Skill launches an expert reviewer agent that reads the project's CLAUDE.md, checks the PR diff against a detailed standards checklist, posts a structured review on GitHub, and files tracked follow-up issues for anything deferred. ## Core Features & Use Cases - Standards-based PR review: Checks bash 3.2 portability, dependency-free Node ESM assets, generated-index freshness, template guard consistency, security issues, and architecture alignment against the repo's own rules. - Structured review output: Produces a severity-tiered review (critical, suggestions, nitpicks) with a verdict, posts it as a PR comment via gh, and reports a scannable summary table. - Follow-up issue lifecycle: Creates labeled GitHub issues for deferred items, links them in the review, and closes resolved from-review issues with a PR reference. - Use Case: After opening a PR that modifies a skill template and its registry index, run the review to verify registry.json was regenerated, guards still match, and scripts remain bash 3.2 compatible before merging. ## Quick Start Ask the agent to review the current branch's pull request and post the findings as a PR comment with follow-up issues for deferred items.

Frequently Asked Questions about agent-review

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

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

Invoke the review with an optional PR number; it defaults to the current branch's PR. The agent reads CLAUDE.md, fetches the PR and diff via the gh CLI, evaluates the changes against a standards checklist, and posts the review as a PR comment.

What does the code review checklist cover?▼

It covers bash 3.2 portability for scripts, zero-dependency Node ESM assets, generated registry.json freshness, skill-guards.json consistency, security issues like injection and credential exposure, testing gates, and architecture alignment with established patterns.

How are deferred review suggestions tracked?▼

Every valid but out-of-scope suggestion becomes a GitHub issue created with the enhancement and from-review labels. The issue URL must be linked in the posted review's Deferred Items table so it stays discoverable from the PR.

Can the review close issues fixed by the PR?▼

Yes. After fixes are committed, the agent lists open from-review issues, comments on each one resolved by the PR with a reference to the PR number, and closes it. Every closed issue must reference a PR; silent closes are not allowed.

When should findings be escalated instead of decided in the review?▼

Escalate only when answers lead to materially different work or the action is irreversible or outward-facing, using a structured question with 2-4 options and a recommendation. Cheap, reversible calls are stated with their assumption directly in the review.