AI for Developers
Agent skill (SKILL.md) Open source Apache-2.0 49.6k

open-code-review

Alibaba's code review CLI: it reads a git diff and leaves line-level comments through an LLM agent.

Install

For this project
$ npx skills add alibaba/open-code-review -a claude-code
For every project
$ npx skills add alibaba/open-code-review -a claude-code -g

Or copy the skill folder into .claude/skills/ (project) or ~/.claude/skills/ (all projects).

For this project
$ npx skills add alibaba/open-code-review -a codex
For every project
$ npx skills add alibaba/open-code-review -a codex -g

Or copy the skill folder into .agents/skills/ (project) or ~/.agents/skills/ (all projects).

For this project
$ npx skills add alibaba/open-code-review -a cursor
For every project
$ npx skills add alibaba/open-code-review -a cursor -g

Or copy the skill folder into .agents/skills/ (project) or ~/.cursor/skills/ (all projects).

For this project
$ npx skills add alibaba/open-code-review -a opencode
For every project
$ npx skills add alibaba/open-code-review -a opencode -g

Or copy the skill folder into .agents/skills/ (project) or ~/.config/opencode/skills/ (all projects).

For this project
$ npx skills add alibaba/open-code-review -a kimi-code-cli
For every project
$ npx skills add alibaba/open-code-review -a kimi-code-cli -g

Or copy the skill folder into .agents/skills/ (project) or ~/.agents/skills/ (all projects).

Install the CLI with npm install -g @alibaba-group/open-code-review, then point it at a model endpoint.

When to use

Use it to review a pull request or a local diff before merging, especially if you want fewer false alarms than a general agent usually gives.

Open Code Review is a command line tool that reviews a git diff with an LLM agent and returns structured comments tied to specific lines. It began as Alibaba's internal code review assistant, which the team says served tens of thousands of developers before it was open sourced. You point it at a model endpoint you configure.

The agent can read whole files, search the codebase and look at other changed files for context. A separate ocr scan command reviews entire files, which helps when you are auditing code that has no useful diff. The project's own benchmark, built from 200 real pull requests, claims higher precision and F1 than a general purpose agent on the same model, at about a ninth of the tokens. It also admits recall is lower, a deliberate trade for less noise.

Related skills

All skills