The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions facilitate the ingestion of untrusted external content, which presents a surface for potential indirect prompt injection attacks.\n
Ingestion points: The references/astra-workflow.md file directs the agent to retrieve and inspect "real reference images" from user-supplied data and external image searches performed via the host.\n
Boundary markers: There are no explicit instructions for using delimiters or boundary markers to isolate instructions from the potentially untrusted image data or search results.\n
Capability inventory: The skill possesses capabilities for executing Blender diagnostic tools (blender_inspect_motion), managing filesystem caches, and initiating rendering sub-workflows.\n
Sanitization: The workflow emphasizes visual inspection for structural accuracy but does not include automated sanitization or filtering of image metadata or search content.\n- [EXTERNAL_DOWNLOADS]: The skill references downloads from established and trusted external entities.\n
The documentation in references/astra-workflow.md suggests the use of Poly Haven for asset acquisition and references official OpenAI documentation for model guidance. These are recognized as safe, well-known services and trusted organizations within the development and 3D communities.\n- [COMMAND_EXECUTION]: The skill instructions involve the use of local system commands for Blender operations.\n
SKILL.md outlines procedures for baking simulations, inspecting motion with the blender_inspect_motion tool, and managing local cache manifests, which requires the agent to execute tools within the host environment.

blender-simulation-workflow