baoyu-url-to-markdown

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads project dependencies via bun install during the setup phase. It fetches content from user-provided URLs and utilizes a remote API (https://defuddle.md) as a fallback for high-quality HTML-to-markdown conversion. Additionally, it downloads images and videos from source pages to local directories if the --download-media flag is enabled.
  • [COMMAND_EXECUTION]: The skill requires the bun runtime to execute its internal TypeScript scrapers. It also uses osascript on macOS to activate the browser window when user interaction is required (e.g., for solving CAPTCHAs or logging in).
  • [CREDENTIALS_UNSAFE]: To support scraping login-protected content (specifically for X/Twitter), the skill manages session cookies. These are saved to and restored from a local JSON file (x-session-cookies.json) located within the user's Chrome profile directory. This behavior is a functional requirement for the skill's stated purpose of handling authenticated content.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process arbitrary, untrusted HTML from the internet. While it converts this data into markdown, the resulting text should be handled with caution by the AI agent, as the source content could theoretically contain instructions intended to influence the agent's downstream behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 11:17 AM