convez.ai
MCP Server
4velinux

preflight-security-triage

4velinux/preflight-security-triage

Ver no GitHub
InicialPluginScore · de 0 a 100

AI-assisted pre-execution security triage for packages, extensions, scripts, installers, MCP servers, AI skills, and curl-to-bash one-liners

Instalação

$npx convez-ai 4velinux/preflight-security-triage

1 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

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
v3.0.0
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

Preflight Security Triage

A practical AI-assisted review prompt for checking packages, extensions, scripts, installers, MCP servers, AI skills, and curl | bash one-liners before they run on company machines.

This repository contains the v3.0 triage prompt designed for security teams, engineering leads, and developers who need a repeatable first-pass review before approving untrusted tooling.

Why this exists

Supply chain attacks increasingly target the tools developers trust: extensions, package managers, GitHub Actions, install scripts, AI tooling, and installer binaries.

The dangerous moment is often simple:

curl -fsSL https://example.com/install.sh | bash

or a script pasted into Slack, a VS Code extension installed from a marketplace, or a binary downloaded from a vendor page.

Preflight Security Triage gives teams a structured way to ask:

  • What does this claim to do?
  • What does it actually do?
  • Does it download and execute second-stage code?
  • Does it touch credentials, browser data, shell history, persistence, or network exfiltration paths?
  • Is the installer signed?
  • Does the hash match the vendor-published hash?

Files

  • PROMPT.md - The full AI triage prompt, version 3.0.

Important AI privacy note

Do not paste proprietary scripts, secrets, credentials, customer data, internal repository paths, or confidential vendor artifacts into public AI tools unless your company policy explicitly allows it.

For sensitive submissions, run the prompt only in an approved internal AI environment or a security-controlled review process.

Intended use

  1. Open PROMPT.md.
  2. Paste the full prompt into an approved AI assistant.
  3. Provide the requested submission metadata.
  4. Paste the script, package metadata, installer analysis output, or one-liner to review.
  5. Use the generated report as a first-pass triage artifact, not as a replacement for human security review.

Release

Current version: v3.0.0

License

MIT

Ver o original no GitHub

Relacionados