product-flow-design

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes JSON data from external Lucid diagrams to summarize workflows or perform edits. This represents a potential surface for indirect prompt injection if diagram text contains malicious instructions.\n
  • Ingestion points: Lucid diagram data fetched via mcp__lucid__fetch (SKILL.md, Section 2).\n
  • Boundary markers: Instructions do not specify delimiters to separate untrusted diagram content from agent instructions.\n
  • Capability inventory: The agent has capabilities to create, modify, and delete elements within the user's Lucid account (SKILL.md, Sections 3 and 4).\n
  • Sanitization: No input sanitization or filtering logic is defined for the content extracted from diagram nodes.\n- [EXTERNAL_DOWNLOADS]: The skill fetches configuration schemas and architectural guidelines from established platforms including Google Drive, Google Docs, and Lucid.app to inform its design decisions (SKILL.md, Section 0).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 07:31 PM
Security Audit — agent-trust-hub — product-flow-design