AIGate

Stop risky AI-generated Git changes before push.

CI OpenSSF Scorecard npm version License: MIT

• README: English | 한국어 | 日本語 | 中文
• Operations guide: English | 한국어 | 日本語 | 中文
• HTML overview: https://leehueeng.github.io/aigate-ai-git-workflow-guard-cli/

AIGate is a zero-config pre-push safety CLI for AI-assisted coding. It checks changed files, possible secrets, repository readiness, PR risk, and branch strategy before changes reach your remote branch or pull request review.

AIGate terminal demo

AIGate demo

• Your AI coding assistant changed files faster than you can review them.
• You want one command that explains readiness before git push.
• You maintain an open source package and need PR, release, and repository health signals.
• You want Markdown, HTML, JSON, and SARIF outputs that work locally and in CI.
• You want Codex, Gemini, and Claude Code to follow the same branch and validation workflow as humans.

Run it without installing:

npx -y aigate-cli check
npx -y aigate-cli start --route default --dry-run
npx -y aigate-cli start --route quickstart --dry-run
npx -y aigate-cli doctor
npx -y aigate-cli test
npx -y aigate-cli ai report
npx -y aigate-cli demo
npx -y aigate-cli pr-check
npx -y aigate-cli evaluate-project

Or install it globally:

npm install -g aigate-cli
aigate start
aigate start --route default --ask-steps
aigate start --route default --steps init,repo-files
aigate start --route oss --dry-run
aigate reset --dry-run
aigate clean
aigate uninstall
aigate check
aigate test
aigate ai report
aigate aitest
aigate git-ready
aigate install-hook --pre-push
aigate pr-check

Try it from source:

git clone https://github.com/LeeHueeng/aigate-ai-git-workflow-guard-cli.git
cd aigate-ai-git-workflow-guard-cli
npm install
npm test
node src/cli.mjs check

AIGate git-ready: READY
Branch: feature/my-work
Changed files: 4
Secret findings: 0
Project score: 92/100
Blockers: none
Warnings: none
Recommendation: Run AIGate test, commit focused changes, push the branch, and open a pull request.

AIGate is useful when AI coding assistants move quickly and you need one repeatable local gate before git push or PR creation.

AIGate is strongest when it is wired into a workflow, not only run as advice. It reports three practical levels:

• advisory: a person can ignore AIGate and run plain git push.
• partial: a hook is active in this clone, but it can still be bypassed with --no-verify and may not be active in a fresh clone.
• server enforced: CI runs aigate git-ready and verified branch protection or merge rules require that CI to pass.

aigate doctor reports the current level as AIGate enforcement. aigate evaluate-project checks both AIGate CI gate exists and AIGate server enforcement exists, so a CI job that only runs in advisory mode does not look fully protected. doctor also separates an active local hook from a committed hook file and hook activation automation, because core.hooksPath is clone-local. For GitLab, AIGate treats allow_failure: true or when: manual jobs as non-enforcing. Declared settings such as --gitlab-pipeline-must-succeed true are reported as declared evidence; server enforcement only passes when evidence is marked verified. When a CI gate exists but verified server enforcement is missing, evaluate-project keeps the raw score in JSON and caps the final score below A-grade so the warning remains visible.

Read the full tool comparison for recommended combinations with Husky, Lefthook, pre-commit, and Gitleaks.

git switch -c feature/my-work
aigate ai report
aigate start --route default --ask-steps
aigate start --route oss --dry-run
aigate start --route ai --provider all
aigate reset --dry-run
aigate clean
aigate doctor
aigate install-hook --pre-push
aigate test
aigate aitest
aigate git-ready
git add <files>
git commit -m "feat: my focused change"
aigate push -u origin feature/my-work
aigate pr-check --output .aigate/reports/pr.md
aigate pr --title "feat: my focused change"
aigate github comment --pr <number>
aigate github check --output .aigate/reports/github-check.md
aigate trends record
aigate github setup --owner @your-org/team --dry-run

Send a Slack notification when a blocker appears:

AIGATE_SLACK_WEBHOOK_URL="https://hooks.slack.com/services/..." \
 aigate git-ready --notify-channel slack

Use AIGate as a reusable public GitHub Action:

name: AIGate
on:
 pull_request:
jobs:
 aigate:
 runs-on: ubuntu-latest
 steps:
 - uses: actions/checkout@v7
 - uses: LeeHueeng/aigate-ai-git-workflow-guard-cli@v0.1.7
 with:
 command: git-ready
 language: en

Use the current release tag for stable runs, or @main only when you intentionally want unreleased behavior. Full inputs are documented in GitHub Action. The same action metadata is mirrored at .github/actions/aigate for local workflow testing.

Action release status:

• Current stable tag: v0.1.7
• Action usage: ready through LeeHueeng/aigate-ai-git-workflow-guard-cli@v0.1.7
• Marketplace publishing: manual GitHub Release screen step
• Action name: AIGate Git Workflow Guard
• Primary category: Code quality
• Secondary category: Security
• Release title: AIGate Git Workflow Guard v0.1.7

Generate repository instructions for Codex, Gemini, and Claude Code:

aigate integrate all
aigate ai report
aigate ai report --apply --provider codex
aigate aitest --provider codex
aigate aitest --apply --provider codex

This creates AGENTS.md, GEMINI.md, CLAUDE.md, and .aigate/integrations/* so AI assistants follow the same branch, validation, and guarded push workflow as human contributors. aigate aitest writes .aigate/reports/ai-test.md; --apply is the explicit switch that lets AIGate invoke Codex, Claude, Gemini, or a custom --agent-command.

aigate ai report combines the current Git state, repository foundation score, release readiness, branch strategy, and AI handoff guidance. It does not edit files by default. Add --apply --provider codex|claude|gemini only when you want AIGate to run the selected AI CLI with the generated project brief.

When --apply runs, AIGate shows the prompt path, provider, agent command, and live agent output in the terminal, then captures stdout/stderr in the report.

AIGate supports English, Korean, Japanese, and Chinese CLI output:

aigate setup --language en
aigate setup --language ko
aigate setup --language ja
aigate setup --language zh

These public install paths are available now:

npm install -g aigate-cli
brew install LeeHueeng/tap/aigate-cli
docker run --rm -v "$PWD:/repo" -w /repo ghcr.io/leehueeng/aigate-cli:0.1.7 check

These remain on the roadmap:

• Standalone binaries
• Hosted dashboard
• Deeper Linear and Jira workflow automation

• Usage guide
• Distribution guide
• Tool comparison
• Notifications guide
• AI integrations
• Basic Node project example
• JSON output examples
• Windows smoke test notes
• Branch strategy
• Release process
• Hotfix process
• Git upload workflow
• Roadmap
• Multilingual documentation index
• Changelog
• English operations guide
• 한국어 운영 문서
• 日本語運用ドキュメント
• 中文运维说明
• Local visual HTML guides: ko, en, ja, zh

• Star or watch the repository if AIGate catches a risky push in your workflow.
• Open a real-world example issue showing how AIGate behaves in your repository.
• Share the demo GIF, terminal screenshot, or assets/social-preview.png when introducing the project.
• Pick a good first issue if you want to improve docs, examples, integrations, or packaging.

Small issues are welcome. Good first contributions include docs examples, GitHub Actions examples, Homebrew/Docker validation notes, and real-world repository case studies.

npm install
npm test
npm run ci

Read CONTRIBUTING.md, SECURITY.md, and CODE_OF_CONDUCT.md before opening a pull request.

MIT