logo-with-variants
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external SVG files to generate JSX components and update registry configuration. This process involves reading untrusted content which could theoretically contain instructions designed to influence the agent's code generation.
- Ingestion points: User-provided SVG files found in
public/test/or other specified paths (referenced in Step 1 and Step 3). - Boundary markers: The instructions do not define specific delimiters or "ignore instructions" markers for the content read from SVGs.
- Capability inventory: The skill performs file writes to
src/components/ui/logos/andregistry/, and executes thebunbuild command. - Sanitization: The skill performs attribute transformation (e.g., converting
fill-ruletofillRule), but does not explicitly describe sanitization of non-attribute content like metadata or comments where malicious instructions could be hidden. - [COMMAND_EXECUTION]: The skill includes a step to run a local build script to finalize the registry update.
- Evidence: Execution of
bun run build:registryin Step 7. This is a standard and expected action within the context of a Node.js/Bun development environment.
Audit Metadata