audit-data-exposure
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a bundled Python script (
scripts/scan_repository_data_exposure.py) to execute Git commands viasubprocess.run. The implementation follows industry best practices for security by utilizing a 'safe' Git environment that explicitly disables global/system configurations, Git LFS smudging, and Git hooks (core.hooksPath=/dev/null). This prevents a malicious repository from triggering code execution during the audit process. - [PROMPT_INJECTION]: The skill is designed to process untrusted data from repositories, which presents a surface for indirect prompt injection. The skill mitigates this by using a scanner that returns redacted JSON metadata rather than raw file content to the agent, and by providing strict instructional boundaries that forbid the agent from reproducing detected values, snippets, or surrounding text in its output.
Audit Metadata