page-agent

Pass

Audited by Gen Agent Trust Hub on Apr 20, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates the creation of agents that interpret natural language commands to interact with web page DOM elements. This creates an indirect prompt injection surface where malicious content on a webpage or malicious user input could potentially influence the agent's actions.
  • Ingestion points: Untrusted data enters the agent context via userInput and input variables in the Vue, React, and native JavaScript examples found in SKILL.md and references/component-templates.md.
  • Boundary markers: While the skill does not show explicit delimiter-based prompt engineering, it provides a security mitigation via the onBeforeAction hook in SKILL.md and references/use-cases.md, allowing for human-in-the-loop verification of actions.
  • Capability inventory: The integrated Page Agent can perform browser actions such as clicking, form filling, and searching as documented in SKILL.md and references/use-cases.md.
  • Sanitization: The provided code snippets do not include explicit sanitization or validation of the natural language inputs before they are passed to the agent.execute() method.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 20, 2026, 09:16 AM