stitch-design-md

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external data from design projects to generate local files.
  • Ingestion points: Project metadata, screen designs, styles, and component details are fetched via the Stitch MCP server from user-supplied URLs or IDs (SKILL.md Step 2 and 3).
  • Boundary markers: The prompt instructions do not include specific delimiters or warnings to ignore malicious instructions that might be embedded within design strings or metadata (e.g., component descriptions).
  • Capability inventory: The agent is instructed to write a DESIGN.md file to the local project root or the design-system/ directory (SKILL.md Output Format).
  • Sanitization: There is no evidence of sanitization, validation, or escaping of the design project data before it is interpolated into the markdown output documentation.
  • [SAFE]: The skill documentation indicates it originates from a trusted organization (Google) and integrates with official Stitch MCP tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 03:01 PM
Security Audit — agent-trust-hub — stitch-design-md