suede-ab-testing
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured guidelines, checklists, and templates for designing and analyzing experiments. It does not perform automated command execution, network exfiltration, or persistence operations.
- [SAFE]: References to external tools (e.g., Optimizely, PostHog, VWO) and calculators (e.g., Evan Miller) point to established and well-known services in the A/B testing industry. No typosquatting or suspicious URL patterns were identified.
- [SAFE]: The skill includes explicit security best practices, such as warning the user never to paste credentials into skill files or prompts, and defines clear boundaries for agent behavior regarding traffic allocation and results reporting.
- [SAFE]: Data ingestion is limited to reading local marketing context files to inform the agent's questioning; the skill lacks any dangerous capabilities (such as shell access or arbitrary file writing) that could be exploited via indirect prompt injection.
Audit Metadata