The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of ffmpeg and ffprobe to perform forensic video analysis, including scene detection, frame extraction, and audio processing. These commands are executed via the shell to process user-provided video files.
[EXTERNAL_DOWNLOADS]: The pipeline invokes npx hyperframes transcribe to handle audio transcription. This command downloads and executes code from the npm registry at runtime. While npm is a well-known service, the package is a third-party dependency not explicitly listed as a trusted vendor.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (video files and their transcripts) using LLM sub-agents, which creates a potential surface for instructions embedded in the video to influence the agent's behavior.
Ingestion points: The system ingests local reference videos (ref.mp4) and user-provided scripts as described in SKILL.md and references/method.md §1.
Boundary markers: There are no explicit delimiters or system instructions provided to the sub-agents (in references/method.md §5) to ignore potential instructions contained within the visual frames or transcribed audio.
Capability inventory: The skill has capabilities to execute shell commands (ffmpeg, ffprobe), write files to the working directory (style-guide.md, build-directives.md), and manage project directories.
Sanitization: No sanitization, filtering, or validation is performed on the extracted transcripts or visual content before they are passed to the vision and text sub-agents for forensic analysis.

reel-style-clone