adversarial-ux-test
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and acts upon untrusted data from external URLs and project documentation to drive its roleplay, feedback generation, and ticket creation activities.
- Ingestion points: Untrusted data enters the agent's context via browser tools navigating to user-provided URLs and by reading project documentation, as instructed in Step 2 of
SKILL.md. - Boundary markers: The skill utilizes a phased workflow and explicitly commands the agent to "Step OUT of the persona" during the pragmatism filter in Step 4, which establishes a cognitive boundary between processing external content and performing assessment tasks.
- Capability inventory: The agent has the capability to navigate web applications, capture screenshots, inspect browser console logs, and generate actionable bug/UX tickets based on its findings (
SKILL.md). - Sanitization: The instructions do not define any sanitization, validation, or filtering of the content retrieved from external websites or project documents before it is processed by the agent.
Audit Metadata