The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from captured websites to generate narration scripts and video compositions.
Ingestion points: Website data is extracted into files like extracted/visible-text.txt and extracted/tokens.json (references/step-1-capture.md).
Boundary markers: Absent; the agent is instructed to read the raw extracted files and incorporate specific numbers and features directly into scripts.
Capability inventory: The skill uses npx hyperframes for various tasks, performs file writes, and generates executable Javascript for compositions.
Sanitization: Absent; the workflow relies on the agent to manually summarize and transcribe the captured text.
[EXTERNAL_DOWNLOADS]: The skill references external assets and tools from well-known and trusted sources.
Fetches animation libraries (Lottie, GSAP) from jsdelivr, a well-known CDN (references/techniques.md).
Downloads typography assets from Google Fonts (references/techniques.md).
Uses npx to execute the hyperframes and tsx packages at runtime.
[DYNAMIC_EXECUTION]: The skill involves generating Javascript-based compositions at runtime.
Generates HTML and GSAP (Javascript) code for video scenes based on the brand identity and captured website assets (references/step-6-build.md). This dynamic generation is the primary purpose of the skill.

website-to-hyperframes