ab-testing
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues or malicious patterns were detected. The skill follows best practices for providing domain-specific instructions to an AI agent.
- [EXTERNAL_DOWNLOADS]: The skill contains links to reputable, well-known external tools such as Evan Miller's calculator, Optimizely, and VWO for sample size and duration calculations. These are strictly for user reference and do not involve automated script downloads or execution.
- [DATA_EXFILTRATION]: The instructions prompt the agent to read local project files like
.agents/product-marketing.mdto gain context. This is expected behavior for an agent skill and does not target sensitive system credentials or private keys. - [PROMPT_INJECTION]: The hypothesis framework and instructional guidelines do not attempt to override system prompts or bypass AI safety constraints. The 'important' markers are used benignly to emphasize experimental rigor.
Audit Metadata