design-data

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to bootstrap an MCP server using @adobe/design-data-mcp via npx.
  • Evidence: The package @adobe/design-data-mcp is a vendor-owned resource associated with the skill author (adobe).
  • Evidence: The skill references a remote rule hosted on Adobe's official GitHub repository: https://github.com/adobe/spectrum-design-data/.
  • [REMOTE_CODE_EXECUTION]: The skill configuration executes a remote package as a command-line tool to provide its core functionality.
  • Evidence: The Bootstrap section provides an MCP configuration using "command": "npx" and "args": ["-y", "@adobe/design-data-mcp"]. This executes the latest version of the Adobe-authored MCP server.
  • [INDIRECT_PROMPT_INJECTION]: The design-data-query and design-data-suggest tools ingest natural language strings from the user context.
  • Evidence: design-data-query accepts a filter string and design-data-suggest accepts an intent string. While these process untrusted user data, the tools are scoped to querying a static design-token snapshot, which limits the potential impact of adversarial input.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 07:56 AM
Security Audit — agent-trust-hub — design-data