plasmic-designer

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill's setup process involves installing the chrome-devtools-mcp package via npx. This is an official tool from the Chrome DevTools team used to bridge the AI agent to the browser. As this originates from a well-known service, it is considered a legitimate dependency for the skill's operation.
  • [DYNAMIC_EXECUTION]: The skill performs dynamic JavaScript execution in the browser using the evaluate_script tool. This is used to interface with the window.PLASMIC_AI_TOOLS object within Plasmic Studio for tool discovery and executing design commands (e.g., identify, read, insertHtml). This behavior is restricted to the browser context of the target application.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface as it reads project structure, component names, and property values from the active browser page.
  • Ingestion points: Data is ingested via the read tool call to window.PLASMIC_AI_TOOLS (referenced in SKILL.md).
  • Boundary markers: None explicitly defined for data read from the browser DOM, though the skill instructions emphasize structural parsing for UUIDs.
  • Capability inventory: The skill can navigate pages, execute JavaScript in the browser, and take screenshots.
  • Sanitization: No specific sanitization or escaping of the ingested web content is described prior to parsing.
  • [DATA_EXFILTRATION]: The skill transmits environment metadata (agent model name and client type) to the Plasmic Studio web application through an identify() call. This is used to initialize the design session and involves only non-sensitive environmental strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 03:05 AM