The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/assemble.py executes system utilities ffmpeg and ffprobe to concatenate media files and process audio. These operations are performed using the subprocess module with arguments passed as a list, which is the recommended practice for mitigating command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input in the form of a 'theme', which is subsequently used to generate video scripts, image generation prompts, and burnt-in subtitles. This flow represents an indirect prompt injection attack surface.
Ingestion Points: User-supplied theme input in SKILL.md.
Boundary Markers: The instructions do not define explicit boundary markers or isolation protocols for the user-provided text during its transition between sub-skills.
Capability Inventory: The skill has the capability to write to the filesystem within the outputs/ directory and execute the ffmpeg binary.
Sanitization: The assemble.py script includes escaping logic for file paths when constructing FFmpeg subtitle filter strings, reducing the potential for path-based injection into the media engine.

auto-short-video