adversarial-ux-test

Pass

Audited by Gen Agent Trust Hub on Apr 27, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill directs the agent to adopt an adversarial, 'grumpy,' and 'sweary' persona. This encourages the agent to bypass standard safety guidelines regarding professional tone and helpfulness within the roleplay context.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it instructs the agent to navigate and process content from untrusted external URLs. Ingestion points: Website content, browser console logs, and external project documentation are read via browser tools. Boundary markers: There are no instructions or delimiters defined to separate untrusted data from the agent's core instructions. Capability inventory: The agent has the ability to navigate URLs, capture screenshots, read console logs, and create tickets/files. Sanitization: No validation or sanitization of content ingested from the browser is required before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 27, 2026, 07:07 AM