setup-interview
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses Git commands to version the configuration file. The commit messages are restricted to pre-defined templates ('fresh' or 'update'), which prevents user-supplied input from being injected into the shell command.
- [DATA_EXFILTRATION]: While the skill collects project metadata (e.g., GitHub username, team size), it does not perform network operations. It includes an explicit consistency check to ensure that sensitive information like service-role keys or secrets are not stored in the configuration file.
- [PROMPT_INJECTION]: The instructions focus on structured data gathering and do not contain patterns intended to bypass safety guidelines or override agent behavior.
- [INDIRECT_PROMPT_INJECTION]: As a setup tool, the skill naturally ingests untrusted data from the user. It mitigates risk by performing consistency checks (e.g., duplicate field validation) and restricting its write capabilities to a single, specific JSON configuration file.
Audit Metadata