Guide · Agent skill
How to use the Luna Code Reviewer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Luna is an agent skill: a folder with a SKILL.md instruction file that teaches an AI coding agent a specific job. This guide shows how to install the Luna Code Reviewer skill, ask for a review in plain language, and read the report it produces. It is written for developers who have not used agent skills before.
Luna Code Reviewersickn33/agentic-awesome-skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the Luna skill ZIP from this page and unzip it. The folder contains SKILL.md.
- 2Choose an install location. Place the luna folder in ~/.claude/skills/ for personal use, or in .claude/skills/ inside a project to share it.
- 3Or upload it in the Claude apps. Go to Settings → Capabilities → Skills and upload the skill there instead.
- 4Name the revision. Tell the agent which commit, branch or pull request to review, and list the relevant file paths.
- 5Ask for a review in plain language. Request a security or reliability review and mention that you want findings ranked by severity.
- 6Read the findings. Work through CRITICAL and HIGH items first, using each finding's file, line and fix.
- 7Verify before merging. Check the suggested fixes yourself and re-run the review after changes.
What the Luna skill makes the agent do
Once installed, the agent reviews a specific code revision against its requirements and a design blueprint. It looks for real defects such as injection flaws, missing auth checks, race conditions, unhandled async errors and missing timeouts.
Each finding is ranked from CRITICAL to LOW and written up with the file, line, risk and a concrete fix. The skill deliberately skips naming, formatting and style comments, which linters handle better.
- Security checks: SQL, NoSQL, command and path traversal injection, authentication and authorization gaps, hardcoded secrets, input validation.
- Reliability checks: unhandled promise rejections, missing DB transactions, read-modify-write races, N+1 queries, timeouts and retries on external calls, unbounded queries.
- Severity levels: CRITICAL, HIGH, MED, LOW, with a PASS, PASS WITH CONDITIONS or BLOCK overall status.
Install the Luna skill from the ZIP
Download the ZIP from this page and unzip it. The package contains a single file, SKILL.md. Place the unzipped luna folder in one of two locations: ~/.claude/skills/ to make it available to you across projects, or .claude/skills/ inside a specific project to share it with that repository.
You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/luna/
- Project install: .claude/skills/luna/ inside your repository
Trigger the skill with natural requests
You do not need a special command. Describe the review you want in ordinary language and name the revision or files, so the agent knows what to read. The skill asks for an exact revision and artifact paths, so include them when you can.
The source describes the skill as scoped to a specific revision against its requirements. Reviews of a vague 'whole project' are less precise than a named commit or pull request.
- Review the changes in commit abc1234 for security problems before I merge.
- Do a Luna-style review of the login and session code in src/auth for missing auth checks.
- Check this pull request for race conditions and unhandled async errors, and rank findings by severity.
Read the structured report
The report starts with a summary count of findings by severity and an overall status. Each finding lists the file, line, what is wrong, the risk if it stays, and a specific fix. Blueprint conformance and checklist results follow, then a handoff recommendation and notes on where extra testing is needed.
Treat CRITICAL and HIGH findings as blockers. The skill's own rule is that CRITICAL items must be fixed before any handoff and HIGH items before QA.
- CRITICAL: exploitable security flaw or data loss risk.
- HIGH: incorrect behavior, crashes or data integrity issues under real conditions.
- MED: edge-case or scale problems, fix before deployment.
- LOW: minor risk or technical debt, deferred.
Safety notes: scripts and network
According to the listing, this skill runs no scripts and uses no network access. It requires nothing beyond the agent reading the code you point it at. The skill's own instructions also ask that private data and secrets be kept out of reports, so do not paste credentials into your request.
The skill is a review aid, not an audit. Its own limitations note says AI agents can hallucinate or give incorrect guidance, so verify findings and generated fixes before shipping.
- Runs scripts: no
- Uses network: no
- Requires: nothing listed
What people use it for
- Second-pass security review of a pull request before merging.
- Checking a specific commit for missing auth checks on protected routes.
- Looking for unhandled async errors and missing timeouts in service code.
- Producing a severity-ranked report for a team that uses AI coding agents.
- A checklist-driven pass for injection, auth and secrets problems on a solo project.
Frequently asked questions
Is the Luna Code Reviewer skill free?
The source lists the skill under the MIT licence. It does not list any price, so check the licence text in the repository for the terms that apply.
What do I need to run the Luna skill?
The listing states no requirements beyond the agent itself. The skill runs no scripts and uses no network access, and it works by reading the code you point it at.
Where do I install the Luna skill?
Unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.
Does Luna check code style and formatting?
No. The skill explicitly excludes naming, formatting and style comments unless they create a real correctness or readability risk. Use a linter for style.
Can Luna replace a human security audit or penetration test?
No. The skill only reviews code the agent can read and cannot guarantee it finds every defect. It is a second review, not a substitute for a human audit.
Do I need the other squad agents like Mason, Aria, Quinn and Dep?
No. The file refers to those roles, but the instructions say they are optional and not installed agents. The skill works on its own by reading the code and any artifacts you provide.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Luna suits developers and teams who want a structured, security-minded second look at a specific pull request or commit, with findings ranked by severity. It is a poor fit if you need a human security audit, a style review, or the multi-agent handoff workflow, since those roles are not included in this file. Pair it with a linter and a proper audit where the stakes are high.
Download the skill