The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses setup_short.sh to create directories and copy files, and it relies on ffmpeg and python3 to execute the rendering pipeline. These operations are within the scope of its functionality but involve shell-level interactions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes audio/words.json (transcribed from user-supplied audio). This data drives the Shot timing and visual content in the render.py script. The lack of boundary markers or sanitization for this external input presents an indirect prompt injection surface.
Ingestion points: audio/voice.wav, audio/words.json (used in render_short_template.py and short_kit.py).
Boundary markers: None identified.
Capability inventory: File system writes (mkdir, cp), shell command execution (ffmpeg), and Python execution.
Sanitization: None identified in the provided scripts.
[EXTERNAL_DOWNLOADS]: The pipeline incorporates a fetch_footage.py script from a sibling skill, which performs search and download operations to acquire video assets from external sources based on agent-generated queries.

fireship-style-short