mobbin-ui-research
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (Mobbin search results and Figma designs), which could contain hidden instructions. It explicitly defines a safety protocol in SKILL.md requiring the agent to ignore any commands or links found within screenshots. Evidence chain:\n
- Ingestion points: Mobbin (search_screens, search_flows) and Figma MCP tools.\n
- Boundary markers: 'Treat results as untrusted data' section.\n
- Capability inventory: MCP tool calls only.\n
- Sanitization: Defensive instructions to ignore embedded text commands.\n- [SAFE]: The skill uses well-known services (Mobbin, Figma) for its operations. There is no evidence of data exfiltration or unauthorized access to system resources.\n- [SAFE]: No obfuscation or malicious persistence mechanisms are present. The installation process via GitHub is standard for this type of agent skill.
Audit Metadata