experiment-design
Pass
Audited by Gen Agent Trust Hub on Feb 18, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [PROMPT_INJECTION] (SAFE): The use of terms like 'CRITICAL' and 'NEVER' is confined to instructional methodology for experiment design (e.g., 'NEVER peek at results'). No attempts to bypass agent safety filters or override system instructions were found.- [EXTERNAL_DOWNLOADS] (SAFE): A link to a GitHub repository is provided in the footer for attribution purposes. There are no automated downloads, package installations, or script executions.- [NO_CODE] (SAFE): This skill consists entirely of Markdown text and methodology guidelines. It does not contain any executable scripts, binary files, or code dependencies.
Audit Metadata