page-agent
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for users to run commands such as
npm install page-agent,git clonefor the official repository, and variousnpmscripts for building and testing the agent locally. - [EXTERNAL_DOWNLOADS]: The instructions include downloading the page-agent package from a public registry and cloning the project repository from Alibaba's official GitHub organization.
- [REMOTE_CODE_EXECUTION]: The skill offers a bookmarklet and a script tag to load and execute the page-agent library directly from a CDN for evaluation purposes.
- [INDIRECT_PROMPT_INJECTION]: The integrated agent reads the page Document Object Model (DOM) as text, which represents a surface for indirect prompt injection from untrusted web content. -- Ingestion points: The agent reads and processes the DOM of the current browser page as input. -- Boundary markers: The skill notes that developers can use allowlist hooks to restrict the agent's operational scope. -- Capability inventory: The agent is capable of driving the UI by clicking elements and filling form fields based on instructions. -- Sanitization: The library supports data-masking hooks to filter sensitive information during processing.
Audit Metadata