The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied assets and references to construct Blender scenes. This creates a surface where malicious instructions could be embedded within external data to influence the agent's output or command execution.
Ingestion points: supplied references, requested assets, and user's idea as defined in SKILL.md.
Boundary markers: The skill requires the creation of an implementation brief and validation evidence for each milestone, which provides structural checks but no strict data sanitization.
Capability inventory: modeling, materials, lighting, animation commands, and arbitrary Python execution within the Blender context.
Sanitization: The instructions include explicit checks to stop on path escape or failed validation, providing basic behavioral guardrails.
[DYNAMIC_EXECUTION]: The skill mentions the use of "expert Python" as a fallback for advanced modeling or animation tasks. This involves generating and executing Python code at runtime within the Blender process to achieve goals not covered by standard registered commands.
[EXTERNAL_DOWNLOADS]: The instructions provide a command to install an optional extension (blender-use) using npx skills add full-aigc-skills/blender-skills. This resource is managed by the same author organization and is used to extend foreground control and recovery capabilities.

blender-design