The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses the Blender command-line interface (e.g., blender --background --python) to automate scene initialization, rendering, and export tasks. The binary path is resolved via the $BLENDER_BIN environment variable or standard installation paths.- [DYNAMIC_EXECUTION]: The documentation suggests the use of exec(open(path).read()) for interactive execution within a live Blender session. Additionally, the skill utilizes a template (build_template.py) to generate phase-specific modeling scripts, which are subsequently executed as local files.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided reference images (concept art, turnarounds, screenshots) which represent an ingestion point for untrusted data into the modeling pipeline. * Ingestion points: Reference images are loaded and processed in scripts/init_master.py and scripts/compose_review.py. * Boundary markers: Not explicitly defined for image data; the agent follows a rigid gate protocol. * Capability inventory: The skill maintains the ability to execute local scripts, write files (renders, JSON reports, .blend files), and perform network operations via potential third-party plugins in the host environment. * Sanitization: Standard image processing libraries (Pillow and Blender's internal image engine) are used for handling external assets.

blender-image-to-3d