caveman-evidence-review
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a specialized CLI tool,
caveman, to fetch structured data such as cost reports, health scores, and trace summaries. The instructions enforce a strict read-only policy, prohibiting the agent from performing state-changing operations like starting or stopping experiments. - [DATA_EXFILTRATION]: The skill accesses cloud-based monitoring and cost data, but the access is scoped to the specific project context. Instructions specifically direct the agent to avoid fetching sensitive request/response payloads by default, focusing instead on metadata and aggregate statistics unless the user explicitly requests otherwise.
Audit Metadata