The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (highlight_cut.py and asr.py) via the command line to perform video analysis, transcription, and cutting. These scripts are part of the skill's internal repository.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted audio content through transcription, which the agent then evaluates to select segments.
Ingestion points: The agent reads the output of asr.py (transcribed video content) from /tmp/hl.json in the 'Way B' workflow.
Boundary markers: There are no explicit instructions or delimiters used to help the agent distinguish between transcribed data and its own operational logic.
Capability inventory: The skill has the ability to execute subprocess commands (python skills/shared/scripts/highlight_cut.py) to manipulate the file system (outputting videos to outputs/).
Sanitization: No validation or sanitization is mentioned for the text content generated by the transcription process before it is analyzed by the agent.

video-highlights