code-review

Reviews GitHub pull requests across a 12-phase workflow covering scope, correctness, and complexity.

Updated Mar 6, 2026
One-click install
npx skills add https://github.com/timi-ty/agent-forge --skill code-review-timi-ty
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/timi-ty/agent-forge/tree/main/skills/code-review
Command: npx skills add https://github.com/timi-ty/agent-forge --skill code-review-timi-ty

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Reviewing pull requests thoroughly is time-consuming and error-prone: reviewers miss scope creep, dead code, vacuous tests, and semantic bugs that compile cleanly. This Skill enforces a disciplined, senior-engineer-level review process so nothing slips through. ## Core Features & Use Cases - 12-phase structured workflow: Gathers PR context via the gh CLI, validates scope against linked issues, studies existing codebase patterns in an isolated git worktree, then reviews file-by-file. - Adversarial semantic verification: A separate re-read of every changed file checks whether tests would actually fail if the feature broke and whether application logic does what it claims. - Cyclomatic complexity audit: Counts decision points per function using defined rules, maps findings to a cause-to-transformation catalog, and reports prioritized fixes. - Use Case: A teammate opens a PR touching 15 files. Run this Skill to produce a prioritized review report (High/Medium/Low), post an approve or request-changes verdict to GitHub after your confirmation, and optionally squash-merge. ## Quick Start Ask the agent to review a pull request by pasting its GitHub URL, for example: review this PR https://github.com/owner/repo/pull/123.

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?

Provide the PR URL, owner/repo#number reference, or PR number, and the Skill fetches metadata and the diff using the gh CLI. It then runs a 12-phase review covering scope, patterns, correctness, and complexity, ending with a prioritized report and verdict.

What does an automated PR code review check for?

It checks scope alignment with the PR title and linked issues, conformance to existing codebase patterns, correctness and edge cases, efficiency, dead code, type safety, security, and cyclomatic complexity. A separate semantic verification pass confirms tests would actually fail if the feature broke.

Does the code review skill require the GitHub CLI?

Yes, it uses the gh CLI to fetch PR metadata, diffs, and linked issues, and to post approve or request-changes reviews. It also uses git worktrees to read base-branch code without touching your working directory.

Can the review skill approve or merge a pull request automatically?

No action is taken without explicit user confirmation. After the review report, you confirm or override the verdict, and merging only happens if you approve and then explicitly confirm the squash-merge prompt.

How is cyclomatic complexity measured during the review?

Each added or modified function is counted starting at 1, adding points for conditionals, loops, switch cases, catch clauses, and boolean operators. Functions exceeding defined thresholds are reported with a dominant cause and a concrete refactoring transformation.

What are the limitations of automated PR review?

The review relies on static diff analysis and cannot guarantee runtime behavior unless you optionally check out the PR head to run builds and tests. It also only flags functions the PR touches, so pre-existing complexity elsewhere is not reported.