baoyu-markdown-to-html

Warn

Audited by Gen Agent Trust Hub on Apr 19, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: Dynamically imports and executes JavaScript modules for syntax highlighting from an external CDN (cdn-doocs.oss-cn-shenzhen.aliyuncs.com) using the import() function at runtime in scripts/vendor/baoyu-md/src/utils/languages.ts.
  • [EXTERNAL_DOWNLOADS]: Fetches and downloads remote images from URLs specified within input markdown files to a local directory for processing, as documented in scripts/vendor/baoyu-md/src/images.ts.
  • [DATA_EXFILTRATION]: Sends markdown content contained within PlantUML code blocks to a remote rendering server (plantuml.com) to generate diagram images, as seen in scripts/vendor/baoyu-md/src/extensions/plantuml.ts.
  • [PROMPT_INJECTION]: Exhibits an indirect prompt injection surface by processing untrusted markdown data without robust sanitization or explicit boundary markers.
    • Ingestion points: Reads user-provided markdown files via the inputPath parameter in scripts/main.ts.
    • Boundary markers: Absent; the skill does not wrap interpolated data in markers or include instructions to ignore embedded commands.
    • Capability inventory: Includes file system read/write access, network operations (HTTP/HTTPS GET and fetch), and dynamic execution of scripts.
    • Sanitization: Lacks a dedicated HTML sanitizer for the main output body; sanitization is only applied to metadata summary fields using basic regular expressions.
  • [COMMAND_EXECUTION]: Instructs the agent to execute shell commands using the bun runtime or npx to perform the markdown-to-html conversion.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 19, 2026, 03:23 PM