The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use various shell commands for routine operations, such as blender -b for headless rendering, uv for managing the execution environment, git clone for documentation access, and diff for verifying add-on installations.
[DYNAMIC_EXECUTION]: It details the usage of an execute_blender_code tool which performs dynamic execution (exec) within the Blender process. The instructions specifically guide the agent to treat this as a last resort and favor safer, structured API tools whenever possible.
[EXTERNAL_DOWNLOADS]: The skill references the official Blender Foundation forge at projects.blender.org for cloning the project wiki and accessing the Gitea API to check releases and issues.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes complex external data including .blend scene files and .rst documentation. It mitigates logic errors by mandating visual verification of rendered output to confirm the state reported by scene measurements.
[DATA_EXFILTRATION]: It describes methods for handling large data payloads by writing JSON files to the local temporary directory (bpy.app.tempdir) and reading them via the shell to bypass socket communication limits.

blender