The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the video-studio-engine package directly from a GitHub repository archive and utilizes the @scrollmark/cli tool. These external resources are managed through standard package managers (pip and npx) and are hosted within the official ecosystem of the skill's author.
[COMMAND_EXECUTION]: Multiple bundled Python utility scripts (doctor.py, normalize_audio.py, poster.py, and qc_render.py) utilize the subprocess module to execute system commands. These operations are strictly directed at media processing utilities like ffmpeg and ffprobe, or the project's own video-studio command-line interface, to perform tasks such as audio normalization, frame extraction, and render validation.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves ingesting and processing data from external and potentially untrusted sources, including user-provided URLs, local files, and generated media prompts. This ingestion creates a theoretical surface for indirect prompt injection where malicious instructions could be embedded in media metadata or source content.
Ingestion points: Step 3 of the pipeline acquires footage and audio from diverse sources including url: and file: paths defined in the storyboard.
Boundary markers: The engine uses a structured JSON schema for storyboards and plans to isolate instructions from data, and requires explicit user approval before installations or major pipeline steps.
Capability inventory: The skill possesses capabilities for file system writes to project directories, shell command execution for media processing, and network access for asset sourcing.
Sanitization: Validation is primarily handled by the underlying video-studio engine and ffmpeg filters, supplemented by mandatory human-in-the-loop review phases (Step 8) for quality control.

video-production