kodevu — Claude Code skill: what it does, how to install
Extracts Git or SVN diffs, submits them to an AI reviewer CLI, and writes code analysis reports to Markdown or JSON.
npx skills add gyteng/kodevu --skill kodevu -a claude-codeWhat kodevu does
This Node.js utility inspects code changes directly from project repositories. It works with local Git folders or SVN checkouts and repository URLs. You can target specific revisions using --rev, evaluate a series of recent commits with --last, or inspect uncommitted workspace modifications via --uncommitted.
Review tasks run against backends such as openai, gemini, codex, or copilot. The default auto mode discovers compatible CLI utilities in your system PATH. Summaries land in ~/.kodevu/ as Markdown documents, but you can request JSON outputs and redirect destinations via --output.
Configuration precedence is strict: CLI flags take priority over environment variables, entries in ~/.kodevu/config.json, and defaults. If the target reviewer binary is missing from PATH and no API keys are supplied, execution stops with an error.
What it helps with
- inspect uncommitted workspace edits before pushing
- produce Markdown or JSON review summaries for recent commits
- feed Git or SVN patch diffs to external model reviewers
- enforce custom review guidelines using prompt instruction files
Example prompts for the agent
- Review the latest three commits in this repository using kodevu
- Run kodevu on uncommitted changes and output JSON reports to ./reports
- Run a code review on the latest commit focusing on security flaws
How to install kodevu
1Skills CLI (Vercel)
npx skills add gyteng/kodevu --skill kodevu -a claude-code2Manually: put the files in place
~/.claude/skills/kodevu/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/kodevu/
1Skills CLI (Vercel)
npx skills add gyteng/kodevu --skill kodevu -a codex2Manually: put the files in place
.agents/skills/kodevu/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add gyteng/kodevu --skill kodevu -a cursor2Manually: put the files in place
.cursor/skills/kodevu/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install gyteng/kodevu kodevu2Manually: put the files in place
.github/skills/kodevu/SKILL.md1Skills CLI (Vercel)
npx skills add gyteng/kodevu --skill kodevu -a gemini-cli2Platform CLI
gemini skills install https://github.com/gyteng/kodevuManually: put the files in place
.devin/skills/kodevu/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add gyteng/kodevu --skill kodevu -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about kodevu
How to install kodevu in Claude Code?
Run npx skills add gyteng/kodevu --skill kodevu -a claude-code to add the skill to your agent environment.
What reviewers can kodevu work with?
It supports openai, gemini, codex, and copilot. The default auto setting scans installed CLI tools in PATH.
Where does kodevu save review reports?
Default Markdown reports go to ~/.kodevu/. You can switch to JSON using --format json or pick another path with --output <dir>.
Source text
SKILL.md · MIT
---
name: kodevu
description: A tool to fetch Git/SVN diffs, send them to an AI reviewer, and generate configurable code review reports.
---
# Kodevu Skill
Kodevu is a Node.js tool that fetches Git commits or SVN revisions, sends the diff to a supported AI reviewer CLI, and writes review results to report files. It supports an optional persistent config file at `~/.kodevu/config.json` for settings that should survive across sessions.
## Usage
Use `npx kodevu` to review a codebase.
### Reviewing the latest commit
```bash
npx kodevu .
```
### Reviewing a specific commit
```bash
npx kodevu . --rev <commit-hash>
```
### Reviewing the last N commits
```bash
npx kodevu . --last 3
```
### Reviewing uncommitted changes
```bash
npx kodevu . --uncommitted
```
### Supported Reviewers
`kodevu` supports several AI reviewer backends: `auto`, `openai`, `gemini`, `codex`, `copilot`. The default is `auto`, which probes available CLI tools in your `PATH`.
Example using OpenAI:
```bash
npx kodevu . --reviewer openai --openai-api-key <YOUR_API_KEY> --openai-model gpt-5-mini
```
### Generating JSON Reports
By default, review reports are generated as Markdown files in `~/.kodevu/`. You can specify `--format json` or change the output directory using `--output <dir>`.
```bash
npx kodevu . --format json --output ./reports
```
### Custom Prompts
You can provide additional instructions to the reviewer using `--prompt`:
```bash
npx kodevu . --prompt "Focus on security issues and suggest optimizations."
```
Or from a file: `--prompt @my-rules.txt`
### Environment Variables
All options can also be set via environment variables to avoid repetitive flags:
- `KODEVU_REVIEWER` – Default reviewer.
- `KODEVU_LANG` – Default output language.
- `KODEVU_OUTPUT_DIR` – Default output directory.
- `KODEVU_PROMPT` – Default prompt instructions.
- `KODEVU_OPENAI_API_KEY` – API key for `openai`.
- `KODEVU_OPENAI_BASE_URL` – Base URL for `openai`.
- `KODEVU_OPENAI_MODEL` – Model for `openai`.
### Configuration File
For persistent settings that survive across shells and AI tool invocations, create `~/.kodevu/config.json`:
```json
{
"reviewer": "openai",
"openaiApiKey": "sk-...",
"openaiBaseUrl": "https://your-gateway.example.com/v1",
"openaiModel": "gpt-4o",
"lang": "zh"
}
```
The file is optional and silently ignored if absent. Priority: **CLI flags > ENV vars > config file > defaults**.
## Working with Target Repositories
- **Git**: `target` must be a local repository or subdirectory.
- **SVN**: `target` can be a working copy path or repository URL.
```bash
npx kodevu /path/to/project --last 1
```
© gyteng · MIT · gyteng/kodevu