The Agent Skills Directory

[PROMPT_INJECTION]: The sub-agent contract in references/frame-worker-core.md includes explicit instructions to 'supersede the skill catalog's own imperatives' and directs agents to 'do not open' specific documentation files (SKILL.md), which is a pattern used to override system-level constraints or prior instructions.
[PROMPT_INJECTION]: The skill's frame packet builder in scripts/lib/frame-packets-core.mjs processes user-controlled markdown files to construct instructions for sub-agents, creating a surface for indirect prompt injection. Ingestion points: STORYBOARD.md and BRIEF.md. Boundary markers: Absent; instructions are defined as being copied 'verbatim' into sub-agent prompts. Capability inventory: File system read/write access, sub-agent dispatch, and local project script execution. Sanitization: No sanitization or escaping of external markdown content is performed before interpolation into prompts.
[EXTERNAL_DOWNLOADS]: The documentation references external script loads for the GSAP animation library and the Tailwind CSS browser runtime via the jsdelivr CDN. These are well-known technology services and are documented here as safe, legitimate dependencies of the video rendering framework.

hyperframes-core