The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses spawnSync to execute ffmpeg and ffprobe for audio processing, looping, and format verification within scripts/assemble-index.mjs and scripts/audio.mjs.
[COMMAND_EXECUTION]: The orchestrator is instructed to run multiple npx hyperframes commands for project initialization, authentication status checks, linting, rendering, and previewing (e.g., npx hyperframes init, npx hyperframes render).
[EXTERNAL_DOWNLOADS]: The skill utilizes npx to execute platform-specific tools (hyperframes) and includes instructions for updating the skill from a remote repository via the skills update command.
[DYNAMIC_EXECUTION]: In scripts/audio.mjs, the audio processing engine can be overridden using the HF_MEDIA_ENGINE environment variable, which dynamically determines the script path executed by spawnSync.
[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary user-supplied text (articles, notes, or topics) to generate narrative storyboards and scripts. This ingestion point constitutes an attack surface, although it is part of the primary intended functionality of the skill.
Ingestion points: capture/extracted/visible-text.txt (via user input), BRIEF.md.
Boundary markers: The instructions recommend using a phrase-segmented voiceover structure which can act as a natural delimiter, though no explicit security boundary markers are enforced on user text.
Capability inventory: Subprocess calls (ffmpeg, ffprobe, npx), file writing (HTML, JSON, MD), and sub-agent dispatch.
Sanitization: Standard template-based generation of HTML and JS via scripts; however, raw text is interpolated into storyboard/script drafts by the agent.

faceless-explainer