responsive-screenshots
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the mcp__playwright__browser_evaluate tool to execute a JavaScript snippet in the browser context. This script is static and performs benign layout measurements, such as calculating scroll width and identifying overlapping elements.- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8).
- Ingestion points: External web content is brought into the agent's context through mcp__playwright__browser_navigate and DOM evaluation in SKILL.md.
- Boundary markers: The instructions do not define clear delimiters or provide warnings to the agent to disregard instructions found within the processed web data.
- Capability inventory: The skill uses mcp__playwright__browser_take_screenshot, which allows writing files to the local file system.
- Sanitization: The layout probe script uses .slice(0, 25) to truncate text content retrieved from elements, which provides a basic level of sanitization by limiting the length of any potential injection payload.
Audit Metadata