The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/render_check.py script executes the openscad binary using subprocess.run to generate 3D meshes and preview images. This execution is performed with a structured argument list to mitigate common command injection risks.
[EXTERNAL_DOWNLOADS]: The skill uses the uv package manager to automatically fetch standard Python dependencies (such as trimesh and numpy) and suggests the optional download of the BOSL2 library from its official GitHub repository for advanced design tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, including text prompts and images of sketches, to derive design parameters.
Ingestion points: User-supplied prompts and sketch images processed in the early stages of the workflow (SKILL.md, Step 2).
Boundary markers: Implements a mandatory 'Design brief gate' (SKILL.md, Step 3) requiring explicit user confirmation of the design specifications before any code is written or executed.
Capability inventory: File system access for writing .scad and .stl files, and the ability to trigger external modeling tools.
Sanitization: The instructions require the agent to transcribe requirements into a structured table and feature tree, which are then reviewed by the user, providing a check against malicious instructions embedded in the source data.

design-scad