browser-visual-ux-qa
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from web pages (via browser simulation and screenshots) and possesses powerful capabilities to modify source code and redeploy applications. This creates a risk where instructions embedded in the UI of a site being audited could influence the agent's code modifications.
- Ingestion points: Browser page content, console logs, and full-page screenshots (SKILL.md Phase 2 and 3).
- Boundary markers: No specific delimiters are used to separate UI content from instructions during the critique phase.
- Capability inventory: Writing to files, executing shell commands to start servers/seed data, and performing git operations (commit/push/deploy) (SKILL.md Phase 0, 4, and 6).
- Sanitization: No explicit sanitization or validation of the content extracted from the browser before it is used to guide code fixes.
- [COMMAND_EXECUTION]: The skill requires the ability to execute various shell commands to manage the development environment and deployment pipeline.
- Evidence: Instructions in Phase 0 and Phase 6 include starting servers, seeding databases, committing code, and deploying to live environments. While these are primary functions of the skill, they represent high-privilege operations.
Audit Metadata