top-design

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to evaluate and refactor web designs which involves ingesting user-provided landing pages and design descriptions. This creates an attack surface where malicious instructions could be embedded in the content being analyzed.
  • Ingestion points: User-provided landing pages and design descriptions processed via the 'Top-Design' scoring and refactoring prompts in SKILL.md.
  • Boundary markers: No specific delimiters or 'ignore instructions' directives are provided to isolate user content.
  • Capability inventory: The skill provides design patterns and reference code (CSS/JS); it does not invoke privileged system tools, executables, or network operations.
  • Sanitization: The instructions do not define filtering or sanitization for embedded prompt injections in external content.
  • [METADATA_POISONING]: There is an inconsistency in the author attribution between the skill context and the metadata.
  • Evidence: The SKILL.md YAML frontmatter and affiliate tags in the 'Further Reading' section reference 'wondelai', while the external distribution context identifies 'emmraan' as the author.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 11:36 AM
Security Audit — agent-trust-hub — top-design