design-judge-skills

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions guide users to install components via npx skills add SeanJ1ang/design-judge-skills, which fetches source code and dependencies from a repository on GitHub.
  • [REMOTE_CODE_EXECUTION]: The installation process involves the execution of remote scripts through the npx package runner at installation time.
  • [COMMAND_EXECUTION]: The documentation encourages running local shell commands for testing (e.g., python tests/test_evaluation.py) and managing skill lifecycle (e.g., npx skills update).
  • [DATA_EXFILTRATION]: The skill performs network operations to scrape official design award websites (e.g., ifdesign.com, red-dot.org) for data verification. It also supports an optional BROWSERLESS_API_KEY environment variable, which directs scraping traffic through an external service.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted user data such as design briefs, research summaries, and technical specifications into the agent's context.
  • Ingestion points: Materials processed by the design-evaluation, design-information-prep, and design-submission-check tools in SKILL.md.
  • Boundary markers: No explicit delimiters or warnings are mentioned to prevent the model from following instructions embedded within user-provided design documents.
  • Capability inventory: The skill possesses network capabilities (scraping) and file system access (processing materials).
  • Sanitization: No sanitization or filtering of external document content is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 12:26 AM
Security Audit — agent-trust-hub — design-judge-skills