next-step
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
analyze.jsscript executes localgitcommands (e.g.,git diff,git status) and internal Node.js scripts to assess the current state of the repository.\n- [PROMPT_INJECTION]: The skill contains instructions for an automated execution mode (--goflag) and processes project data that could host indirect injections. \n - Ingestion points:
git statusoutput and markdown files indocs/features/.\n - Boundary markers: Data is summarized into JSON findings by
analyze.jsbefore being presented to the agent.\n - Capability inventory: Subprocess execution of
git/nodeand agent-levelSkilltool invocation.\n - Sanitization: Action arguments are filtered by the script to include only valid file paths and flags.\n- [DATA_EXFILTRATION]: The skill accesses repository metadata and local documentation files within the
docs/features/directory to track project progress and identify missing reviews or documentation updates.\n- [REMOTE_CODE_EXECUTION]: The helper script uses dynamic path resolution andrequirecalls to load internal utility modules from the local environment.
Audit Metadata