analytic-workbench
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill uses directive language in its description to guide agent tool selection for analytic tasks. This is standard functional instruction for skill routing and does not target safety filter bypass.
- [EXTERNAL_DOWNLOADS]: The reference guides recommend installing established data science and orchestration packages such as marimo, DVC, and Hydra. These are trusted external dependencies used for their intended analytic purposes.
- [COMMAND_EXECUTION]: The workbench operates by executing Python scripts and CLI tools. The provided code templates and workflow guides demonstrate standard, legitimate usage for analysis and experiment tracking without evidence of malicious intent.
- [DATA_EXFILTRATION]: No exfiltration patterns were detected. The skill actively promotes security best practices by advising the use of environment variables (.env files) for credential management instead of hardcoding secrets.
Audit Metadata