Open Code Review skill
A skill and CLI for code review: a team of reviewer agents inspects changes independently, debates findings and produces one report
- 382 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Installation changes the project's assistant configuration and adds blocks to AGENTS.md and CLAUDE.md
- On a separate command it posts comments to a pull request in your name through the GitHub CLI
- The fix mode launches an agent that edits the code
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/spencermarx-open-code-reviewDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Installation changes the project's assistant configuration and adds blocks to AGENTS.md and CLAUDE.md
- On a separate command it posts comments to a pull request in your name through the GitHub CLI
- The fix mode launches an agent that edits the code
This is third-party code. Review the repository files before installing.
What it does
Open Code Review, OCR for short, installs as the npm package `@open-code-review/cli`. The `ocr init` command finds the AI assistants installed on your system and writes the ocr skill and slash commands for each. 14 tools are supported, among them Claude Code, Cursor, Windsurf, Codex, Gemini CLI, Cline and GitHub Copilot. A review starts with `/ocr:review` in Claude Code and Cursor or `/ocr-review` in Windsurf and other clients. A lead agent, the Tech Lead, studies the `git diff`, picks a team of reviewers and runs an eight-phase process. Reviewers go through the code in parallel and independently, then discuss the findings: they confirm, challenge, connect and raise new questions. The result arrives as one prioritized report. The library holds 28 reviewer personas in four tiers: generalists, specialists, reviewers in the manner of well-known engineers and your own. Team composition and the model for each reviewer are set in `.ocr/config.yaml` or with the `--team` flag. You can hand reviewers requirements, a spec, a proposal or a ticket, and the result shows what is met and where the gaps are. For large changes of 20 files or more, `/ocr-map` builds a review map with a dependency graph. The `ocr dashboard` panel opens results in the browser, and `/ocr-post` publishes the report to a pull request through the GitHub CLI.
Who it is for. For developers and teams who write code with AI assistants and want a multi-angle review before the human one: security, architecture, quality and tests.
Good fit when
- Before a pull request you want the changes checked from several angles rather than in one model pass
- The project is driven by a spec and the review must check code against requirements from a file or ticket
- A change touches more than 20 files and a reviewer needs a map: where to start and what depends on what
- For one review you want a stronger model on the sensitive part and a cheaper one elsewhere
- The team wants an agent to go through the findings after a review and propose a fix for each
Not a fit when
- A one- or two-line fix: running several reviewers and a discussion for it is excessive
- You need a replacement for human review: the author states OCR only eases the load on people and catches problems earlier
- An old Node.js is installed: the tool requires version 22.5 or newer
- You want a review with no AI assistant connected: an assistant such as Claude Code or OpenCode runs the reviewers
Example request
Review my staged changes with a team of two security specialists and two generalists and check the code against the requirements in openspec/spec.md.Limitations
You need Node.js 22.5 or newer, Git and one of the 14 supported assistants. Publishing to a pull request requires an installed and authorized GitHub CLI and an open pull request on the current branch. The dashboard fetches models for reviewers only from Claude Code and OpenCode.
The `ocr state` command is called at every phase transition of a review: without the installed CLI the skill does not work, and the README stresses it. Session state lives in `.ocr/` and in Node.js's built-in SQLite database.
Several reviewers and a discussion stage spend more tokens than a single pass, and the README gives no exact figure. After updating the package run `ocr update` to sync skills and commands. The `.ocr/config.yaml` file and your personas are preserved.
How to disable. Remove the package with `npm uninstall -g @open-code-review/cli`, then delete the `.ocr` folder from the project and the slash commands added to assistant directories such as `.claude` or `.cursor`. The OCR blocks in AGENTS.md and CLAUDE.md are marked as managed and can be cut out whole.
FAQ
How do I install Open Code Review in Claude Code?
Run `npm install -g @open-code-review/cli`, then `ocr init` in the project directory: it finds Claude Code and adds the skill and slash commands. A review starts with `/ocr:review`, and `ocr doctor` verifies the install.
How does OCR differ from simply asking an assistant to check the code?
Instead of one pass, several reviewers with different priorities work, then debate the findings. Team composition is configurable, and requirements from a spec or ticket go to all reviewers at once.
Which assistants does Open Code Review support?
Fourteen: Amazon Q Developer, Augment, Claude Code, Cline, Codex, Continue, Cursor, Gemini CLI, GitHub Copilot, Kilo Code, OpenCode, Qoder, RooCode and Windsurf. The `ocr init` command detects them by their config directories.
Can different reviewers use different models?
Yes. `.ocr/config.yaml` holds model aliases and the team composition: a fast model for a generalist and a strong one for a specialist. For a single review the roster and models can be overridden from the dashboard or with the `--team` flag.
Does OCR post the report to GitHub on its own?
No, only on the `/ocr-post` command or a dashboard button. Two modes exist: the team result as is, or a version rewritten in one person's voice following Google's review guidelines, with preview and editing. An authorized `gh` and an open pull request are required.
Discussion
No questions or reviews yet.
Related
SkillsA code-review CLI and plugin for agents: reads a git diff and writes line-level comments with context
cross-review sends plans and code to a second model for review: Codex, Claude Code or Antigravity checks your agent's work.
A Claude Code plugin: Claude does the task, then a stop hook starts an independent Codex review and Claude fixes the findings in the same session
A CLI and three skills that work through review-bot comments on a PR: assess findings, fix code, reply and wait for checks
A skill for second-agent review of plans or code: Claude Code starts rounds, Codex reviews, and they exchange files
kyy-logs code-review-skill
code-review-skill
A code review skill with four independent probes, evidence-backed findings and a merge gate derived from JSON
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Open Code Review (this card) | no | 382 | ||
| open-code-review | no | 49,561 | Alibaba's similarly named CLI: one agent writes line comments on the diff, with no reviewer team | |
| code-review-skill | no | 2,204 | a single skill with language and framework checklists, no reviewer team | |
| code-review-graph | no | not required | 32,612 | a local code graph that serves an agent the context of an edit over MCP; it runs no review |
README in short
The README goes from a quick start to the dashboard, then to features: multi-agent review, multi-model teams, review maps, discourse, repeat rounds and personas. Then come configuration, a command reference, updating and session storage in the `.ocr/sessions` folder. One explicit caveat: OCR does not replace human review.
README badge
Original SKILL.md
--- name: ocr description: | AI-powered multi-agent code review. Simulates a team of Principal Engineers reviewing code from different perspectives. Use when asked to review code, check a PR, analyze changes, or perform code review. license: Apache-2.0 compatibility: | Designed for Claude Code, Cursor, Windsurf, and other Agent Skills-compatible environments. Requires git. Optional: gh CLI for GitHub integration. metadata: author: spencermarx version: "2.5.0" # double quotes required — automated sync via nx release repository: https://github.com/spencermarx/open-code-review --- # Open Code Review You are the **Tech Lead** orchestrating a multi-agent code review. Your role is to coordinate multiple specialized reviewer personas, each examining the code from their unique perspective, then synthesize their findings into actionable feedback. ## When to Use This Skill Activate when the user: - Asks to "review my code" or "review these changes" - Mentions "code review", "PR review", or "check my implementation" - Wants feedback on code quality, security, architecture, or testing - Asks to analyze a commit, branch, or pull request ## ⚠️ IMPORTANT: Setup Guard (Run First!) **Before ANY OCR operation**, you MUST validate that OCR is properly set up: 1. **Read and execute `references/setup-guard.md`** 2. If setup validation fails → STOP and show the user the error message 3. If setup validation passes → Proceed with the requested operation This prevents confusing errors and ensures users know how to fix setup issues. ## Quick Start For immediate review of staged changes: 1. **Run the setup guard** (see above - this is mandatory!) 2. Read `references/workflow.md` for the complete 8-phase process 3. Begin with Phase 1: Context Discovery 4. Follow each phase sequentially