favicon-pack-maker
Pass
Audited by Gen Agent Trust Hub on Jun 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill's instructions focus on legitimate design and repository maintenance tasks. No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill specifically includes a security warning about manifest visibility in admin environments.
- [EXTERNAL_DOWNLOADS]: The skill recommends using standard Node.js libraries such as
sharpandpng-to-icofor local image processing. These are well-known, established packages for this specific use case. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface. 1. Ingestion points: Brand direction and tokens are read from files like
DESIGN.mdand.agents/brand-page-context.md(SKILL.md). 2. Boundary markers: Explicit boundary markers for untrusted documentation content are absent. 3. Capability inventory: The skill writes image files to thepublic/directory and updates metadata in frontend layouts or HTML files (SKILL.md). 4. Sanitization: The skill does not describe specific sanitization for text ingested from documentation. The risk remains low as actions are limited to standard asset generation and metadata updates.
Audit Metadata