svg-designer-acrazie
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external SVG markup and font files which could contain malicious payloads.
- Ingestion points: The skill ingests user-supplied SVG references and font files for icon and banner creation.
- Boundary markers: The skill body (SKILL.md) and references (exports.md) explicitly instruct the agent to treat these files as 'untrusted data' and 'data, not instructions'.
- Capability inventory: Capabilities include writing to the filesystem (SVG and exported raster assets) and potential invocation of command-line rendering tools.
- Sanitization: Instructions require the agent to inspect generated code and specifically forbid implicit network fetches or the execution of code embedded within untrusted source assets.
- [REMOTE_CODE_EXECUTION]: Although the skill references external rendering tools such as Inkscape, Sharp, and librsvg, it explicitly prohibits their installation or execution without separate, verified user approval, specifically identifying that tools like 'npx' should not be used to bypass these restrictions.
- [DATA_EXFILTRATION]: The skill contains explicit negative constraints against exposing credentials, weakening Content Security Policies, or initiating unauthorized network operations to external services.
Audit Metadata