The Agent Skills Directory

[COMMAND_EXECUTION]: Multiple scripts, including content_review.py, join_takes.py, and run_takes.py, use subprocess.run() to invoke system utilities like ffmpeg and ffprobe. These calls are essential for the skill's purpose of processing video and audio files.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating untrusted user data into prompts for AI generation models.
Ingestion points: Character traits such as age, ethnicity, wardrobe, and scene description are ingested via CLI arguments in make_character.py and beats data in plan_takes.py.
Boundary markers: Prompt templates in scripts/character-prompt.json and scripts/plan_takes.py (the TEMPLATE variable) do not utilize distinct delimiters or specific instructions to ignore embedded commands within the user-provided data.
Capability inventory: The skill can perform file system writes and initiate image/video generation requests through external API proxies.
Sanitization: Basic sanitization is present, such as checking for empty inputs in make_character.py and checking for brackets (used for model instructions) in spoken dialogue in plan_takes.py, but it does not comprehensively escape potential injection characters.
[EXTERNAL_DOWNLOADS]: The media_proxy.py script facilitates communication with external AI providers like FAL and ElevenLabs to fetch generated media assets. These downloads are directed to well-known service providers and are consistent with the skill's stated purpose.
[CREDENTIALS_UNSAFE]: The skill manages authentication for its proxy services by reading from ~/.gooseworks/credentials.json. This is the documented credential management mechanism for this utility's environment.

create-creator-takes-h3