The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process video transcripts that may contain spoken instructions from the video creator. The instructions explicitly state that these spoken cues, such as 'remove this' or 'another take', should 'outrank' the AI's judgment. This creates a scenario where untrusted content from the processed video clips can control the agent's output and file-writing behavior.
Ingestion points: Transcripts are read from takes-<id>.txt files generated by local transcription tools.
Boundary markers: The skill does not define specific delimiters to separate the raw transcribed content from the instructional analysis performed by the agent.
Capability inventory: The agent generates the cuts.json file which controls the video rendering engine and can execute several local Python scripts.
Sanitization: There is no evidence of sanitization or filtering of the transcript text before it is analyzed for instructions.
[COMMAND_EXECUTION]: The skill relies on the execution of multiple local scripts located in the tools/ directory, as well as system utilities like ffmpeg and ffprobe. It also starts a local web server for the editor/server.py component.

clean-cut