generate-sandbox-policy
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed as a security control to help users generate restrictive sandbox policies. It includes an explicit validation and warning step (Step 6) that flags broad permissions like L4-only access, full HTTP methods, and wildcard binary paths.
- [EXTERNAL_DOWNLOADS]: The skill fetches REST API documentation for well-known services (such as GitHub, Anthropic, OpenAI, and NVIDIA) to generate more precise policies. These operations target trusted organizations and well-known services, which is consistent with the skill's primary purpose.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of external API documentation. This is a known attack surface (vulnerability vector 8c), but it is mitigated by the skill's internal logic which requires human-in-the-loop clarification (Step 2) and performs structural validation against a fixed schema before generating output.
- [PRIVILEGE_ESCALATION]: The skill includes explicit checks to prevent privilege escalation, such as rejecting the root user (UID 0) in process identity configurations.
Audit Metadata