hallmark

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references several well-known design and asset providers for icons, logos, and placeholders. These include Google Fonts, Simple Icons, SVGL, Lucide, Unsplash, and Pexels. It also references the author's project domain (usehallmark.com) for a curated imagery kit.
  • [PROMPT_INJECTION]: The hallmark study <URL> verb creates an ingestion point for untrusted data from remote web pages. The skill instructions in references/study.md include extensive security guards, explicitly directing the agent to treat fetched HTML/CSS as untrusted inert data and to ignore any remote instructions found in comments, metadata, or visible copy (Indirect Prompt Injection mitigation). It also defines a 'Remote URL safety' protocol that forbids access to local/internal network targets and non-web schemes (SSRF protection).
  • [SAFE]: The skill performs standard development environment operations, such as reading project configuration files (package.json, tailwind.config.js) and maintaining a build history log in .hallmark/log.json. These actions are necessary for the skill's primary function of ensuring design consistency and variety in a codebase.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 03:54 PM
Security Audit — agent-trust-hub — hallmark