browser-intent
Pass
Audited by Gen Agent Trust Hub on Jul 17, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates autonomous browser interactions based on natural language intent and processed web content, creating a risk of indirect prompt injection.
- Ingestion points: The skill accepts a user-provided task description and retrieves DOM data from specified URLs in the SKILL.md file.
- Boundary markers: Safety checks are implemented via AIDefence tools (e.g., mcp__claude-flow__aidefence_is_safe) to identify PII or threat patterns.
- Capability inventory: The agent has access to browser controls and potentially the host system via the allowed Bash and Read tools.
- Sanitization: Output is redacted upon detection of malicious content or sensitive data exposure via the contentFlagged attribute.
- [EXTERNAL_DOWNLOADS]: Fetches the page-agent package from the npm registry, which is an official library provided by Alibaba.
Audit Metadata