tech-stack
Warn
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill facilitates the setup of the
agent-browsertool from Vercel Labs, which includes theagent_browser_evalcapability. This allows the agent to execute JavaScript within a browser environment to assist with debugging and interactive exploration tasks. - [INDIRECT_PROMPT_INJECTION]: By utilizing the
agent-browserMCP for exploring live web applications, the agent interacts with external, potentially untrusted web content. This content serves as an ingestion point for instructions that might influence the agent's subsequent actions. - Ingestion points: Data and snapshots retrieved from external websites via the
agent-browsertool (e.g.,agent_browser_snapshot). - Boundary markers: The skill does not define specific delimiters or instructions for the agent when processing data returned from the browser tool.
- Capability inventory: The skill provides tools for browser navigation, interaction (
click,fill), and JavaScript execution (eval) as described in the documentation. - Sanitization: Content retrieved from the browser is processed by the agent without an intermediate sanitization or filtering layer.
Audit Metadata