skills/drolu/agent-skills/hallmark/Gen Agent Trust Hub

hallmark

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The hallmark study command utilizes the WebFetch tool to retrieve HTML and CSS from user-provided URLs for design analysis.
  • Fetches design metadata such as font declarations and color tokens from remote public websites.
  • References design assets and iconography from established third-party services including Unsplash, Pexels, and Google Fonts.
  • Retrieves curated illustrative assets from the skill's official asset domain, usehallmark.com.
  • [SAFE]: The skill incorporates extensive security rails for its network operations to prevent SSRF and internal network exposure.
  • Explicitly blocks access to local hostnames and private IP address ranges, including 127.0.0.1, 10.0.0.0/8, and the AWS metadata service IP 169.254.169.254.
  • Restricts network fetching to http and https protocols, specifically refusing dangerous schemes like file:, ftp:, and javascript:.
  • [SAFE]: The skill instructions include robust defenses against indirect prompt injection when processing content from external websites.
  • Mandates that all fetched remote content be treated as untrusted data and strictly parsed only for design facts.
  • Explicitly directs the AI agent to ignore instructions, comments, or metadata found within remote HTML and CSS code.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 07:11 PM
Security Audit — agent-trust-hub — hallmark