The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses execSync in scripts/generate-voiceover.ts and scripts/preflight.ts to invoke ffmpeg, ffprobe, and whisper for audio manipulation and transcription. These commands are constructed using parameters from a configuration file generated during the workflow.
[INDIRECT_PROMPT_INJECTION]: The scripts/discover-brand.ts script extracts product names, descriptions, and taglines from a target repository's README.md and package.json. This external data is used to pre-populate prompts for the agent, creating a surface for indirect prompt injection if a repository contains malicious instructions.
Ingestion points: discover-brand.ts reads package.json, README.md, and manifest.json from the target project path.
Boundary markers: The skill implements a multi-phase interactive workflow where the user can confirm or override all discovered data via JSON-formatted questions.
Capability inventory: The skill has the ability to execute shell commands (execSync), write files (writeFileSync), and perform network requests (fetch).
Sanitization: Discovered text is used as-is in the agent's context without explicit filtering or escaping.
[EXTERNAL_DOWNLOADS]: The skill communicates with the ElevenLabs API (api.elevenlabs.io) for voiceover generation and discovery. It also suggests installing openai-whisper via pip and lucide-react via npm, which are standard, well-known packages.

promo-video