page-agent
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the page-agent source code and library dependencies from trusted and well-known sources, including the official Alibaba GitHub repository, the npm registry, and the jsDelivr CDN.
- [COMMAND_EXECUTION]: Provides instructions for running standard development environment setup and build commands (e.g., npm install, npm run build) and for using bookmarklets to inject scripts into the browser for testing.
- [PROMPT_INJECTION]: The skill implements an agent that reads and interprets webpage DOM content to execute instructions, which presents a surface for indirect prompt injection where an attacker-controlled page could attempt to subvert the agent's behavior.
- Ingestion points: Reads webpage DOM content as text for processing (SKILL.md).
- Boundary markers: Documents configuration options for allowlists and data-masking hooks to restrict the agent's scope (SKILL.md).
- Capability inventory: Capable of performing UI interactions such as clicking elements and filling form fields based on natural language instructions (SKILL.md).
- Sanitization: Includes support for developer-defined security hooks to mask data and validate interactions (SKILL.md).
Audit Metadata