Agent Powerups ships instructions, metadata, and helper scripts. The risk surface comes from what an agent can do after loading them.
- Review assets before enabling them.
- Do not trust compatibility claims blindly.
- Do not run install commands or scripts without understanding their effect.
- Give agents least privilege.
This repo should not contain:
- tokens, API keys, or credentials
- machine-specific paths
- internal project names, non-public dataset/table identifiers, or company-specific examples
- personal or customer data
- destructive-only automation
- instructions to bypass security controls
- Hooks may execute code.
- MCP configs may expand tool access.
- Skills may direct an agent to read local files or run commands.
- Install commands can modify the local environment.
More detail: docs/security-model.md
If you find sensitive content or dangerous instructions, please report it privately via GitHub Security Advisories or email the maintainers directly. Do not post sensitive material in public issues.