The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/film.py script utilizes subprocess.run and subprocess.check_output to execute system binaries like blender, ffmpeg, and ffprobe. These executions are driven by user-supplied configuration files, which could lead to unintended command execution if the configuration is malicious. The commands use a list format for arguments, which reduces but does not eliminate risks associated with external tool invocation.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from 3D assets and JSON manifests, creating an attack surface for indirect prompt injection.
Ingestion points: 3D assets are read by scripts/inventory.py; configuration and material manifests are loaded via json.loads in scripts/apply_materials.py and scripts/film.py.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore potential commands embedded within the metadata or material names of processed assets.
Capability inventory: The skill has significant local capabilities, including file system modification, network downloads via curl, and binary execution via subprocess.
Sanitization: Validation logic is present for data structure integrity, but lacks deep sanitization of string content before it is used in prompts or CLI arguments.
[EXTERNAL_DOWNLOADS]: The skill uses curl -fL to download 3D assets and textures from URLs provided by the Scenario MCP server. These operations are essential to the skill's functionality and target the vendor's own infrastructure.
[DYNAMIC_EXECUTION]: The provided Python scripts dynamically modify the sys.path to import local helper modules and use the bpy API to manipulate 3D scenes within the Blender environment. This allows for runtime assembly of 3D processing logic.

scenario-patina-retexture