The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied narrative text from STORYBOARD.md and BRIEF.md to guide sub-agents in generating executable HTML and GSAP animation code.
Ingestion points: scripts/lib/frame-packets-core.mjs reads the project's STORYBOARD.md file to bundle context for sub-agents.
Boundary markers: Data is structured in Markdown sections, but lacks specific escaping or directives to prevent sub-agents from following instructions embedded within user-provided narrative fields.
Capability inventory: Sub-agents generate files (writeFileSync) that are executed in a browser environment.
Sanitization: There is no evidence of validation or sanitization of narrative text before it is presented to sub-agents for code generation.
[PROMPT_INJECTION]: Documentation in references/frame-worker-core.md contains instructions for sub-agents to "supersede the skill catalog's own imperatives" and to "do not open" other instruction files, which matches patterns used to override platform-level behavioral constraints.
[COMMAND_EXECUTION]: The framework's build and management processes involve executing local Node.js scripts for packet generation (scripts/lib/frame-packets-core.mjs) and preference handling (prefs.mjs).
[EXTERNAL_DOWNLOADS]: The composition templates fetch legitimate JavaScript libraries, including GSAP and the Tailwind browser runtime, from the JSDelivr CDN (cdn.jsdelivr.net), which is a well-known service.

hyperframes-core