Skip to content

Security: MontrealAI/skillos

Security

SECURITY.md

Security Policy

Agent SkillOS is a reference implementation. It is designed to demonstrate safe skill learning patterns, not to replace enterprise security controls.

Core safety expectations

  • Do not let agents silently publish global skill changes.
  • Do not let private data become a shared network skill.
  • Do not grant tool access through skill text alone.
  • Require approval for high-impact actions.
  • Keep release history and rollback paths.

Reporting issues

For a real deployment, route security reports through your organization's vulnerability disclosure process.

There aren't any published security advisories