baoyu-url-to-markdown

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The file scripts/lib/browser/cookie-sidecar.ts implements a mechanism to save and restore sensitive session cookies (such as auth_token and ct0 for X/Twitter) to a local JSON file named x-session-cookies.json within the Chrome profile directory. While intended for session persistence between runs, this stores credentials in plaintext on the local filesystem.
  • [EXTERNAL_DOWNLOADS]: The skill utilizes https://defuddle.md as a remote conversion service. In scripts/lib/extract/html-to-markdown.ts, the skill sends the user-provided target URL to this third-party endpoint to fetch converted markdown. Users should be aware that the content of accessed pages is shared with this service.
  • [COMMAND_EXECUTION]: The skill executes several local commands to manage the browsing environment:
  • ps aux is used in scripts/lib/browser/profile.ts to detect existing Chrome processes using specific profiles.
  • osascript is used in scripts/lib/browser/session.ts on macOS platforms to bring the browser window to the foreground during interactive modes.
  • bun install is executed during the initial setup to install runtime dependencies.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external websites (HTML, tweets, transcripts). Because the resulting markdown is intended to be read and potentially acted upon by an AI agent, malicious instructions embedded in a target webpage could attempt to influence the agent's behavior. The skill includes basic sanitization (removing scripts and styles), but the risk of indirect injection remains inherent to web processing tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:47 AM
Security Audit — agent-trust-hub — baoyu-url-to-markdown