better-your-harness
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The scanner script invokes standard command-line tools such as
gitandgh(GitHub CLI) to retrieve repository metadata. These calls utilize argument lists rather than shell string interpolation, which effectively prevents command injection risks from malicious file or directory names. - [PROMPT_INJECTION]: The skill processes project documentation and code that could contain adversarial instructions designed to influence the agent's behavior. The skill mitigates this by using a deterministic scanner to produce facts and explicitly instructing the AI agent to only interpret these facts within fixed boundaries, preventing the agent from following external project-based instructions.
- [SAFE]: The implementation handles sensitive data with caution. While it scans for the presence of credentials (like API keys and tokens) to alert the user, the script is hardcoded to extract only metadata (file paths and variable names) and strictly excludes the actual values of the secrets from all outputs.
Audit Metadata