npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

ai-repo-safety

v0.1.4

Published

AI/vibe-coding repo safety skill with secret scanning, SAST, GitHub read guard, MCP safety, and incident cleanup.

Downloads

416

Vulnerabilities

Powered byPowered by Snyk
  • 0High
  • 0Medium
  • 0Low

Links

Git

Maintainers

artemletyaartemletya

Keywords

securityai-codingvibe-codingskillclaude-codeagentssecretssastmcpgithub

Readme

AI Repo Safety Skill

Package links:

One powerful public skill + asset repository for hardening AI/vibe-coded projects before commit, before push, and before publishing on GitHub.

The project is designed for Python 3.12, uv, and uvx, and works on Windows, macOS, and Linux. It uses only free / open-source / community tools by default.

What this gives you

  • one installable skill: SKILL.md
  • one Python CLI: ai-repo-safety
  • safe repo bootstrap before the first commit
  • secret file denylist
  • Gitleaks / TruffleHog / detect-secrets integration
  • Opengrep-first SAST profile, without Semgrep as a default dependency
  • Python hardening via Bandit, Ruff, pip-audit, pytest, pydantic-settings examples
  • GitHub public repo hardening workflows
  • GitHub read guard for commits, PRs, branches, issues, and merge request aliases
  • MCP config safety checks
  • lightweight STRIDE threat model templates
  • incident cleanup templates
  • cross-platform tool doctor and install plan

Note: Earlier versions of this README referenced skill/ai-repo-safety/SKILL.md. The skill ships as a single SKILL.md at the repository root, and agents are expected to install it into their config directory via the skills CLI.

Install the skill (AI agents)

Install to all detected agents in one command:

npx skills add letya999/ai-repo-safety-skill

Or install to a specific agent:

npx skills add letya999/ai-repo-safety-skill -a claude-code

Branch and release model: the default branch is dev. Releases are tagged from dev (e.g. v0.1.4) and published to PyPI and npm via Trusted Publishing on tag push. The ai-repo-safety verify-release --version X.Y.Z command checks that a release is ready before you push the tag.

Install the CLI

Package pages:

Via uv/uvx (recommended):

uv tool install ai-repo-safety
ai-repo-safety doctor

Or run without installing:

uvx ai-repo-safety doctor
uvx ai-repo-safety init --target . --python auto --github auto
uvx ai-repo-safety scan --target .

Via pip:

pip install ai-repo-safety
ai-repo-safety doctor

Via npm (delegates to Python under the hood):

npm install -g ai-repo-safety
ai-repo-safety doctor

Skill install layout

The skill lives here:

SKILL.md

The skill is ready to be executed via uv run universally on Windows, macOS, and Linux without the need for OS-specific shell wrappers.

Commands

# Read-only environment check.
ai-repo-safety doctor

# Plan-only bootstrap. By default does not install tools, hooks,
# or call the GitHub API. Use --apply --yes and the matching
# opt-in flag to perform a specific mutation.
ai-repo-safety init --target . --python auto --github auto
ai-repo-safety setup --target .            # plan only
ai-repo-safety setup --target . --apply --run-hooks --yes

# Local hook install. Refuses to overwrite an unmanaged existing
# hook unless --overwrite (or --chain to append) is passed.
ai-repo-safety install-hooks --target .
ai-repo-safety install-hooks --target . --chain
ai-repo-safety install-hooks --target . --overwrite

# Scans.
ai-repo-safety scan --target .
ai-repo-safety scan --target . --strict
ai-repo-safety prepush --target .

# GitHub read guard. Always pass an explicit --reason.
ai-repo-safety github-guard validate --target . --repo owner/repo --resource pulls --reason "review current PRs"
ai-repo-safety github-guard read --target . --repo owner/repo --resource pulls --reason "review current PRs"
ai-repo-safety github-guard check-text --target . --file suspicious_issue.md

# Threat model and incident templates.
ai-repo-safety threat-model --target .
ai-repo-safety incident --target . --type secret-leak

# Pre-release verification.
ai-repo-safety verify-release --version 0.1.4 --target .

AI Assistant Integrations

To ensure your AI assistants (like Claude Code, Codex, OpenCode, and Cursor) follow these repository safety rules, you can integrate this skill using the following steps:

1. Installing via skills CLI

If you use a skills manager or custom CLI tool for orchestrating agent abilities, install the skill directly:

skills add git+https://github.com/letya999/ai-repo-safety-skill

This will place the SKILL.md and related guardrails into your agent workspace.

2. Integration with AI Assistants & IDEs

Claude Code (by Anthropic)

Claude Code automatically scans and respects repository instructions. To make it aware of this safety skill:

  1. Place AGENTS.md in the root of your project directory.
  2. When starting a session, Claude Code reads root markdown instructions (like AGENTS.md) and strictly adheres to the forbidden actions and GitHub read guard policies.
  3. You can also reference the CLI directly in your prompt to enforce checks, e.g., claude "run ai-repo-safety scan before committing".

Codex CLI & OpenCode

For CLI-based agents:

  1. Inject the rules by importing the skill or placing the SKILL.md in your agent's config folder.
  2. The agent will read SKILL.md as part of its system instructions, preventing it from performing direct git push or reading raw GitHub API responses without using ai-repo-safety github-guard.

Cursor (IDE)

Cursor uses .cursorrules to guide its Chat and Composer features:

  1. Create a .cursorrules file in the root of your repository (if it doesn't exist yet).
  2. Copy the content of AGENTS.md into your .cursorrules file or append a reference:
    Always follow the repository safety guardrails defined in AGENTS.md.
Never run forbidden actions (like git push, public PR creation) without user confirmation.
  3. Cursor's AI will automatically prioritize these instructions during code generation and terminal executions.

Tool philosophy

Default tools are free / OSS / community:

  • pre-commit
  • gitleaks
  • trufflehog
  • detect-secrets
  • opengrep
  • bandit
  • ruff
  • pip-audit
  • osv-scanner
  • cyclonedx-py
  • Renovate
  • OpenSSF Scorecard
  • optional CodeQL for public/open-source GitHub repos

Semgrep is not the default. The SAST profile is Opengrep-first. Existing Semgrep-compatible rules can be adapted by the agent when compatible.

Tool installation policy

The CLI checks whether Git, Python, uv, uvx, GitHub CLI, and scanners are available.

When tools are missing, the agent can automatically install all required Python and System binaries by running:

uv run ai-repo-safety install-missing

This command uses uv tool for Python tools (like bandit, pip-audit) and leverages the native package manager (winget, brew, npm) or direct downloads to globally install system binaries (like Gitleaks, OSV-Scanner, TruffleHog, Opengrep, GitHub CLI) across Windows, macOS, and Linux.

GitHub read guard

Agents often ingest too much GitHub context: commits, PRs, branches, issues, comments, and public issue bodies with prompt injection. This project includes a read guard:

ai-repo-safety github-guard read --repo owner/repo --resource issues --reason "triage current issues"

It enforces policy from .repo-safety.json:

  • only allowed repositories by default
  • explicit reason required
  • max items
  • max body characters
  • secret redaction
  • prompt-injection pattern detection
  • aliases for mrs / merge_requests -> GitHub pulls

Agents should use this wrapper instead of direct gh api, gh pr view, gh issue view, or raw GitHub web reads when reading GitHub context into an AI session.

Important limits

This project gives strong local deterministic gates, but it cannot magically intercept every external AI tool. The included AGENTS.md and hook templates force agents to use the guardrails, but each client has different hook/plugin support.

The safe default is: if the agent cannot enforce a guardrail in its runtime, it must run the CLI guard before the risky action.