beauty-json

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill uses emphatic and mandatory language ("ABSOLUTELY PROHIBITED", "严禁") to override the AI's standard behavior regarding token limits and summarization, forcing 100% data fidelity during conversion. \n- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through its file ingestion process. \n
  • Ingestion points: The skill reads HTML files from the beauty-html/assets/ directory. \n
  • Boundary markers: There are no specified delimiters or instructions to ignore embedded commands within the input HTML files. \n
  • Capability inventory: The skill's capabilities are restricted to text parsing and file templating; it does not utilize shell execution or direct network access. \n
  • Sanitization: The instructions do not define any sanitization or validation steps for the content extracted from the template files. \n- [EXTERNAL_DOWNLOADS]: The generated HTML output references the Chart.js library via the JSDelivr CDN, which is a well-known and trusted source for hosting open-source libraries.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 03:55 PM