The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted user-provided prompts and instructions which are passed to an external SVG generation model, creating an indirect prompt injection surface.
Ingestion points: The prompt positional argument and the --instructions, --references, and --image flags in scripts/generate.ts accept untrusted external input.
Boundary markers: No explicit boundary markers or 'ignore' instructions are used to delimit user input from system instructions in the provided script logic.
Capability inventory: The skill utilizes the writeFile capability in scripts/generate.ts to save output to the filesystem.
Sanitization: The outputPath variable, derived from the user-controlled --output flag, is not validated or restricted to a specific directory, potentially allowing for arbitrary file write or path traversal if the environment permissions are not constrained.

generate-svg