huashu-design-html-native

Warn

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The installation process uses npx skills add alchaincyf/huashu-design, which downloads and installs code from an unverified third-party repository.
  • [COMMAND_EXECUTION]: The skill invokes multiple local shell and Node.js scripts such as render-video.js, convert-formats.sh, and add-music.sh to perform media encoding, video rendering via Playwright, and PDF/PPTX exports.
  • [PROMPT_INJECTION]: The Brand Asset Protocol introduces a surface for indirect prompt injection.
  • Ingestion points: The agent is instructed to search for and download HTML and SVG assets from arbitrary external brand websites (e.g., brand.com/press) in SKILL.md.
  • Boundary markers: Absent; there are no instructions or delimiters to isolate the agent from malicious instructions potentially embedded in the downloaded third-party HTML/SVG source.
  • Capability inventory: The skill can execute shell scripts, run Node.js code, and automate browser actions via Playwright (SKILL.md).
  • Sanitization: Absent; while the agent is told to grep for hex codes, it still processes the full content of untrusted external files without sanitization.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 21, 2026, 07:41 AM