The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on several shell commands to manage the video creation workflow, including environment checks and rendering.
Evidence: Executes npx hyperframes (doctor, init, lint, validate, render), ffmpeg for frame extraction, and ffprobe for file analysis.
[EXTERNAL_DOWNLOADS]: Fetches external resources required for the video style and typography.
Evidence: Downloads WOFF2 font files from cdn.jsdelivr.net and design configuration metadata from hyperframes.dev.
[PROMPT_INJECTION]: The skill ingests untrusted subtitle data which directly influences the agent's planning and the final video output.
Ingestion points: audio/narration.srt (processed by scripts/srt-cues.mjs and read by the agent).
Boundary markers: Absent; the agent is instructed to read the SRT to plan scenes and generate HTML content without delimiters or safety warnings.
Capability inventory: Execution of shell commands (npx, ffmpeg, node), file system modifications (mkdir, cp), and network operations (curl).
Sanitization: There is no validation or escaping logic to prevent malicious instructions embedded in the SRT files from affecting the agent's behavior.

producing-video