The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill executes Node.js scripts from paths computed at runtime. For example, it calls node <MEDIA_DIR>/scripts/prefs.mjs and node <MEDIA_DIR>/scripts/recipe.mjs, where <MEDIA_DIR> is defined as the directory of a separate installed skill.
[DYNAMIC_EXECUTION]: The skill dynamically constructs worker roles by concatenating instruction files at runtime. The file scripts/frame-packets.mjs facilitates the creation of a _role.md payload which merges core contracts with skill-specific instructions for sub-agents.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data that could influence the agent's logic or composition output. \n
Ingestion points: Figma URLs, BRIEF.md, and STORYBOARD.md project files. \n
Boundary markers: The skill instructs the agent to use specific canonical terms and follow the 'brief contract' to manage checkpoints. \n
Capability inventory: Shell command execution via npx, Node.js script execution, file writing, and worker dispatching. \n
Sanitization: The instructions explicitly state that raw Figma connector calls are avoided to ensure SVG sanitization and brand token binding.
[EXTERNAL_DOWNLOADS]: The skill makes extensive use of npx to execute the hyperframes CLI for skill updates, project initialization, and build validation. These downloads originate from the vendor's ecosystem (HeyGen, Inc.).

hyperframes-general-video