The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a specific Python module (common.runners.cli.model3d) which interfaces with the 3D generation service. This is a standard pattern for modality-specific runners.
[EXTERNAL_DOWNLOADS]: Through the --image-url flag, the skill can be instructed to fetch reference images from external URLs to serve as the basis for 3D mesh generation.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it ingests untrusted user text and remote image data that are subsequently processed by the agent and passed to a shell command.
Ingestion points: User-provided text descriptions and remote image URLs (SKILL.md).
Boundary markers: The command template suggests wrapping user prompts in double quotes, though no explicit sanitization logic is provided within the skill files.
Capability inventory: The skill is granted Bash, Read, Write, and Edit permissions to manage files and execute the runner.
Sanitization: No specific input validation or sanitization routines are defined; the skill relies on the agent's default handling of tool arguments.

model-maker