browser-use
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill defines patterns for browsing and extracting data from external websites, which creates a surface for indirect prompt injection where untrusted content could attempt to influence the agent behavior.
- Ingestion points: Web extraction logic in
SKILL.md(e.g.,page.textContent,page.$$eval). - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are included in the provided code snippets.
- Capability inventory: The skill provides capabilities for browser navigation, interaction, and content extraction.
- Sanitization: No sanitization or validation of the content retrieved from external web pages is demonstrated in the implementation patterns.
Audit Metadata