Skip to content

Security: YuLab-SMU/Rho

SECURITY.md

Security policy

Report suspected vulnerabilities through GitHub private vulnerability reporting. Include the affected source revision or artifact, platform, impact, reproduction steps and minimal evidence. Keep credentials, private project content and unpatched exploit details out of public reports. Maintainers coordinate validation and disclosure through the private advisory; no fixed response deadline is promised.

Scope and current status

Rho owns the application container and its selected composition. Application implementation and official distribution are pending. Design requirements are not evidence that a protection is implemented. Report component-specific issues to Rho-core or Rho-plugins as appropriate. See Status, Privacy and Artifact trust.

Security boundaries

Native programs run with the Host's operating-system authority. A Core caller namespace does not by itself authenticate a principal or sandbox native code. Component identity and resource references must be checked against the selected caller and project scope; models and retrieved content cannot expand that scope. An Agent's existing authorization requires mechanical boundary checks, without another application approval for the same action.

Application-delivered UI must separate trusted native capabilities from isolated external content. Generated interfaces use approved components and action/resource references; schema validation does not establish authorization, resource freshness or scientific correctness. Signing proves an artifact's signing identity and byte integrity, not the safety or correctness of its behavior.

Acceptance must exercise the actual selected application and components, including failure and unknown outcomes. Keep signing and publication credentials out of build/test jobs. Keep secrets out of source and public diagnostics. Distribution controls follow Release.

There aren't any published security advisories