The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a local utility script, scripts/measure.py, which invokes the ffprobe command-line tool.
The script uses subprocess.run with a list of arguments to call ffprobe, which is a secure method of executing external binaries that prevents shell injection.
The utility is used specifically for measuring audio duration and extracting metadata for project synchronization.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external text data such as voice transcripts, dictated chat, and user-provided scripts, which introduces an indirect prompt injection surface.
The skill instructions explicitly direct the agent to treat transcripts as 'evidence, not authority' and to perform manual context checks to resolve recognition errors.
The use of  and  boundary markers in Markdown files helps the agent distinguish between spoken content and potential instruction blocks.
[DYNAMIC_EXECUTION]: The scripts/measure.py script performs runtime analysis of files within the project environment.
It reads file contents to calculate SHA-256 hashes for integrity receipts and uses the wave library or ffprobe for audio inspection.
The script includes input validation and error handling for file paths and numerical values.

video-scriptwriter