visual-testing
Pass
Audited by Gen Agent Trust Hub on Feb 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes bash commands through the
browser-devtools-clitool to perform browser automation tasks, such as navigating to web pages, capturing screenshots, and extracting DOM structures. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its inherent function of ingesting data from untrusted external websites.
- Ingestion points: External data enters the agent's context through tools that retrieve content from URLs, specifically the
navigation go-to,get-as-html, andget-as-textoperations defined inSKILL.md. - Boundary markers: The instructions do not define delimiters or provide specific prompts to the agent to disregard potential instructions embedded within the retrieved HTML or text content.
- Capability inventory: The agent possesses capabilities to manipulate browser sessions, capture screen data, and interact with the Figma API through the provided CLI.
- Sanitization: There are no specified mechanisms for filtering or sanitizing the content obtained from remote web sources before it is processed.
Audit Metadata