Skip to content

Security: ImL1s/aigate

Security

SECURITY.md

Security Policy

Reporting a Vulnerability

Do NOT open a public GitHub issue for security vulnerabilities.

If you discover a security vulnerability in aigate, please report it responsibly:

  1. Email: Send a detailed report to the maintainers via GitHub private vulnerability reporting.
  2. Include:
    • Description of the vulnerability
    • Steps to reproduce
    • Potential impact
    • Suggested fix (if any)

We will acknowledge your report within 48 hours and aim to provide a fix or mitigation plan within 7 days.

Scope

The following are in scope for security reports:

Area Examples
Pre-filter bypass A malicious package that evades all static checks
AI prompt injection Crafted package code that manipulates AI model output to return "safe"
Code execution Any path where aigate executes package code instead of just reading it
Credential exposure API keys, tokens, or sensitive data leaked in logs or output
Dependency vulnerabilities Vulnerabilities in aigate's own dependencies
GitHub Action security Action input injection, secret leakage, or privilege escalation

Out of Scope

  • Detection accuracy (false positives / false negatives) -- please use the false positive template instead
  • Feature requests
  • Issues in third-party AI model behavior (Claude, Gemini, Ollama)

Supported Versions

Version Supported
Latest on main Yes
PyPI releases Yes
Older releases Best effort

Security Design Principles

aigate is built with these security principles:

  1. Never execute package code -- the resolver only downloads and extracts source archives
  2. Fail-open by design -- if aigate crashes, package installation proceeds (safety over availability)
  3. No secrets in config -- AI backends use CLI tools that manage their own auth
  4. Minimal permissions -- the GitHub Action only needs read access to the repository

Acknowledgments

We appreciate the security research community's efforts in keeping the software supply chain safe. Reporters who follow responsible disclosure will be credited in release notes (unless they prefer anonymity).

There aren't any published security advisories