The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to call FFmpeg and FFprobe for all media processing tasks. These operations are essential for its functionality as a video editor and are performed using safe, list-based argument passing to mitigate command injection risks.
[EXTERNAL_DOWNLOADS]: Legitimate network requests are performed to set up fonts and retrieve map tiles. Assets are fetched from trusted domains, including github.com for typography and openstreetmap.org for mapping data, which are required for the skill's graphic features.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted media and configuration data to generate project plans.
Ingestion points: EDIT.json configuration and audio transcripts.
Boundary markers: None identified in the processing logic.
Capability inventory: Subprocess execution for FFmpeg and file system access for rendering.
Sanitization: Numeric values are clamped and basic JSON structure is validated, though the system is designed to render user-provided text into visual output.

vlog-edit