PromptGuard is a local static prompt auditor. It does not send prompt contents to a remote service.
Please open a private security advisory if the repository is hosted on GitHub, or contact the maintainers directly.
Do not include secrets in public issues.
Security-sensitive areas:
- prompt injection and instruction/data boundary rules
- pre-write guard bypasses
- unsafe high-stakes advice rules
- agent adapter installation scripts