The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external .blend files and motion capture libraries, which serve as untrusted data ingestion points. The skill acknowledges this risk and provides explicit security guardrails.
Ingestion points: External .blend files and mocap libraries referenced in SKILL.md.
Boundary markers: Present; the skill instructs using blender --background with auto-execution disabled.
Capability inventory: execute_command (blender CLI), write_file, read_file, edit_file, and create_file (SKILL.md).
Sanitization: Present; instructions mandate disabling embedded driver evaluation and script auto-execution for downloaded assets.
[COMMAND_EXECUTION]: The skill utilizes the blender command-line interface to execute local Python scripts for rig diagnostics. This is a core functional requirement and is documented with appropriate security precautions.

inspect-blender-rigs-motion