landing-page-design

Warn

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation encourages the installation of the belt CLI tool (belt-sh/cli) and additional skills from the inference-sh organization via npx. These sources are not recognized as trusted organizations, classifying them as unverifiable third-party dependencies.
  • [INDIRECT_PROMPT_INJECTION]: The skill templates facilitate passing user-controlled content into inputs for external AI model calls and search assistants without using boundary markers or sanitization. This allows for potential injection where malicious user input could manipulate downstream tool actions.
  • Ingestion points: User-provided text for landing page descriptions, hero image prompts, and search queries for competitor research (found in SKILL.md).
  • Boundary markers: None; example inputs are directly interpolated into JSON structures for CLI execution.
  • Capability inventory: Shell command execution via the belt CLI tool for remote image generation and web-based data retrieval.
  • Sanitization: No validation or sanitization of input strings is demonstrated or required by the instructions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 21, 2026, 08:14 PM
Security Audit — agent-trust-hub — landing-page-design