ar-global-pr-reviewer

Reviews any GitHub pull request remotely by cloning the repo, running tests, and posting inline findings.

1|2|Updated Jul 11, 2026
One-click install
npx skills add https://github.com/mahsanamin/agentic-repos --skill ar-global-pr-reviewer-mahsanamin
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ar-global-pr-reviewer
Source: https://github.com/mahsanamin/agentic-repos/tree/main/skills/ar-global-pr-reviewer
Command: npx skills add https://github.com/mahsanamin/agentic-repos --skill ar-global-pr-reviewer-mahsanamin

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires gh, git, jq.

What problem does it solve? Reviewing a pull request normally requires having the target project checked out locally with its environment configured. This Skill lets you review any GitHub PR from anywhere on your machine: it clones or reuses the repo, pins the exact PR head commit, runs a best-effort build and test pass, and posts a deduplicated inline review grounded in the project's own rules. ## Core Features & Use Cases - Checkout-free PR review: Accepts a PR URL or owner/repo#N, clones under a global cache keyed by owner/repo, and creates a review worktree pinned to the immutable head SHA. - Project-aware review scope: Detects rule files (CLAUDE.md, AGENTS.md, docs/ai-rules, .cursorrules) and existing automated reviewers (CodeRabbit, SonarQube, linters, CodeQL) to skip categories bots already cover and focus on logical bugs, edge cases, and missing tests. - Evidence-backed findings: Runs setup and tests in the background (Gradle, Maven, Node), classifies each finding as PR-introduced, PR-exposed, or pre-existing, checks author intent, and dedups against existing PR comments before posting. - Use Case: A teammate asks you to review https://github.com/your-org/example-service/pull/356 but you don't have that repo locally. Invoke the Skill with the URL; it clones the service, runs its tests, reviews the diff against the repo's rules, and posts an inline review with suggested fixes. ## Quick Start Ask the AI to run ar-global-pr-reviewer with a full GitHub pull request URL, for example global review https://github.com/your-org/example-service/pull/356.

Frequently Asked Questions about ar-global-pr-reviewer

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

FAQPage Schema
How do I review a GitHub PR without cloning the repo locally?▼

Invoke the Skill with a full PR URL or owner/repo#N form. It clones the repository into a global cache under ~/ar-global-pr-reviewer/repos/, creates a review worktree pinned to the PR head commit, and performs the review there.

How does automated PR review avoid duplicating CodeRabbit or SonarQube comments?▼

The Skill scans for .coderabbit.yaml, workflow references to sonarqube, linters, and codeql, then skips categories those tools already cover, such as style and known security patterns. It also fetches existing PR comments via the GitHub API and dedups before posting.

Does the PR reviewer run the project's tests before reviewing?▼

Yes, it attempts a best-effort setup and test run for Gradle, Maven, and Node projects, launched in the background so diff reading proceeds in parallel. The test verdict is logged and folded into the review as one signal among several.

What tools are required to run a global PR review?▼

The gh CLI must be installed and authenticated, along with git and jq. The worktree helper optionally comes from agentic-devkit, but the Skill can create the review worktree inline if that helper is unavailable.

Why does the review abort with a merge-base or PR-moved error?▼

The Skill pins the PR head via refs/pull/N/head and aborts if the ref moves mid-run or if no merge-base exists with the base branch even after deepening the shallow clone. This prevents reviewing a blend of two commits or producing an empty diff that reads as no changes.

Can I review my own pull request with this Skill?▼

Yes, but the Skill detects self-reviews by comparing the authenticated gh user to the PR author and adjusts its behavior. It lowers the threshold for raising small issues and reframes the review prompt toward what a skeptical teammate would ask.