interaction-design-board

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to run a local Python build script and a design server utility to generate and host prototypes. These are standard developer operations performed within the local session workspace.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied HTML and JSON feedback, which presents a surface for indirect prompt injection. This is mitigated by a strong security boundary: prototypes are executed within a sandboxed srcdoc iframe using a restrictive Content Security Policy (default-src 'none') that blocks network access. Mandatory Evidence Chain: 1. Ingestion points: board.json and variant HTML files. 2. Boundary markers: Sandbox srcdoc with strict CSP. 3. Capability inventory: Local script execution and design server hosting. 4. Sanitization: The builder script validates that prototypes are self-contained and do not escape the workspace directory.
  • [COMMAND_EXECUTION]: The test suite includes a secure use of subprocess.run to execute the builder script with specific arguments for automated verification.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 02:11 PM
Security Audit — agent-trust-hub — interaction-design-board