save-md
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust, transparent system for saving content as Markdown. It provides specific instructions for handling various hosts (GitHub, Google Docs, YouTube, etc.) using legitimate endpoints.
- [EXTERNAL_DOWNLOADS]: The skill uses
curlandyt-dlpto fetch content. These are standard tools for the described purpose. Thereferences/source-endpoints.mdfile correctly identifies official endpoints for services like Google Drive and GitHub, including necessary flags like-Lfor following redirects. - [DATA_EXFILTRATION]: While the skill performs network operations to fetch data, it lacks patterns for sensitive file access or credential harvesting. It specifically warns against sending private or internal URLs to third-party services like Firecrawl.
- [COMMAND_EXECUTION]: The skill uses shell commands for content fetching (
curl) and conversion (pandoc,soffice,yt-dlp). These are executed within the scope of the agent's task to save a named source and do not involve arbitrary command injection or privilege escalation. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to handle untrusted data (web pages, PDFs). It mitigates injection risks by instructing the agent to fetch raw bytes and perform direct extraction rather than relying on model summaries. The use of structured YAML frontmatter for provenance further separates metadata from content.
- [OBFUSCATION]: No obfuscated URLs, encoded commands, or hidden text patterns were found in the instructions or evaluation files.
Audit Metadata