The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads pre-trained machine learning weights for speech synthesis (Kokoro), transcription (Whisper), and segmentation (U2Net) to ~/.cache/hyperframes/. These are expected operations for a media processing utility.
[COMMAND_EXECUTION]: Uses npx hyperframes to run CLI commands for asset generation. This involves executing code from the npm registry.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transcribes audio/video and processes external JSON/subtitles, the results of which are returned to the agent's context.
Ingestion points: Reads script.txt, audio.mp3, video.mp4, subtitles.srt, subtitles.vtt, and openai-response.json.
Boundary markers: The skill instructions do not specify any delimiters or warnings to ignore instructions embedded in the processed data.
Capability inventory: The skill can execute shell commands (npx) and install packages (pip), which could be targeted by an injection attack.
Sanitization: No explicit sanitization or filtering of transcribed text is performed before it is handled by the agent.

hyperframes-media