The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install a required toolchain from the vendor's official GitHub repository using pip. Evidence: pip install 'video-studio-engine[generate,vision] @ https://github.com/scrollmark/social-skills/archive/refs/heads/master.tar.gz' in SKILL.md. These resources originate from the identified vendor.
[REMOTE_CODE_EXECUTION]: The installation process for the toolchain (pip install) involves downloading and executing setup code from the remote archive. Evidence: SKILL.md identifies this as a dependency for the Boil and PointerPopups formats.
[DYNAMIC_EXECUTION]: The Boil format supports a plugin-like mechanism where custom Python functions are defined in a local file (boil_shapes.py) and executed by the drawing engine. Evidence: references/formats/boil.md describes this system, noting that the engine injects specific helpers and enforces constraints (no imports, no top-level code, no eval/open) to restrict the execution environment to drawing logic.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input during "Interview" rounds to structure video composition and render parameters.
Ingestion points: User input provided during the interactive interview rounds defined in each format file (e.g., references/formats/explainer.md Rounds 1-3).
Boundary markers: None identified; instructions do not explicitly warn the model to ignore injection attempts within user-supplied content.
Capability inventory: The skill uses the video-studio CLI to generate assets, track motion, and render video scenes.
Sanitization: None identified in the instructional content.

video-formats