The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a multi-step agent pipeline (orchestrator -> frame packets -> sub-agents) where user-influenced data from storyboard and brief documents is translated into executable HTML and JavaScript compositions.
Ingestion points: Untrusted data is ingested from STORYBOARD.md, BRIEF.md, and project-level design tokens in frame.md.
Boundary markers: The dispatch mechanism uses structured markdown packets with explicit headers (e.g., ## Project inputs, ## Assigned storyboard block) to define the context for sub-agents.
Capability inventory: The skill involves writing and assembling HTML/JS files within the project directory and executing render commands via hyperframes-cli.
Sanitization: The framework includes manual checkpoint gates (collaborative mode) and automated npx hyperframes lint and check commands that validate composition logic, timing, and security constraints (like media_crossorigin_breaks_preview) before final rendering.
[EXTERNAL_DOWNLOADS]: The composition templates and reference documentation recommend loading the GSAP animation library and the Tailwind CSS browser runtime from the jsDelivr CDN. These are well-known and established services for delivering front-end web development resources.
[DYNAMIC_EXECUTION]: The frame-packets-core.mjs library and associated workflow scripts programmatically generate composition code and agent instructions at runtime. This behavior is essential to the skill's primary purpose of building automated video renders based on text plans.

hyperframes-core