agt-policy-authoring
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection Considerations]: The skill is designed to analyze repository content, which could potentially contain malicious instructions. To mitigate this, the instructions explicitly command the agent to treat all external data as untrusted and to ignore any conflicting instructions found within the repository.
- [Local Command Execution]: The skill utilizes the
agt-copilotCLI for policy validation. This is a specific, vendor-provided tool used for a well-defined task (schema validation) and does not involve the execution of arbitrary or remote code. - [Credential Safety]: Strong protections are in place to prevent the agent from accessing sensitive files such as environment variables, private keys, or cloud credentials during the discovery phase.
- [Authorization and Persistence]: The skill follows a safe-by-default approach, requiring explicit user confirmation before any policy file is activated or environment variables are modified.
Audit Metadata