icons
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests SVG data from project files to establish design specifications. While this creates a surface for indirect prompt injection via malicious icon metadata, the skill incorporates defense-in-depth measures. Ingestion points: Reads SVG markup from project folders as described in SKILL.md. Boundary markers: Uses a generated icon-spec.md as a reference and defines explicit linting rules to validate outputs. Capability inventory: Involves reading local files, writing new SVG components, and rendering icons in a local HTML tool. Sanitization: The included visualization tool (sheet.html) utilizes a strict whitelist of allowed SVG tags and filters out potentially dangerous elements like script tags.
- [EXTERNAL_DOWNLOADS]: The visualization tool loads typography from Google Fonts. This is a well-known and trusted service, recognized as a safe external dependency for providing UI assets.
Audit Metadata