The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs child_process.spawn within scripts/gen.mjs and scripts/video.mjs to orchestrate local tools such as the dlazy CLI and ffmpeg. These executions are integral to its function as a generation wrapper and media processor.\n- [EXTERNAL_DOWNLOADS]: The persist logic in scripts/gen.mjs and asBase64 in scripts/lib/providers.mjs perform network fetches for remote assets (images and videos). These are saved locally or passed to AI provider APIs as part of the content generation workflow.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it ingests configuration from brand.yaml and storyboard JSON files, which are then interpolated into LLM prompts.\n
Ingestion points: External visual identity configurations and storyboard definitions are ingested via scripts/brand.mjs and scripts/video.mjs.\n
Boundary markers: The skill appends branding constraints to the main prompt via string concatenation without specialized safety delimiters.\n
Capability inventory: The scripts possess the ability to perform file system writes and network operations across multiple AI provider endpoints.\n
Sanitization: Input processing relies on standard JSON and a custom minimal YAML parser without dedicated prompt injection filtering.

ugc-testimonial