The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data from STORYBOARD.md and BRIEF.md, which is then distributed to worker agents through generated instruction packets. This creates a surface for indirect prompt injection where malicious instructions could influence sub-agent behavior.
Ingestion points: The skill reads user-controlled data from STORYBOARD.md and BRIEF.md within SKILL.md and scripts/lib/frame-packets-core.mjs.
Boundary markers: The orchestration script scripts/lib/frame-packets-core.mjs uses markdown headers to separate storyboard content in generated packets, providing basic structure but not full isolation.
Capability inventory: The skill executes shell commands (npx hyperframes), writes files to the project directory, and performs network-based searches.
Sanitization: No explicit sanitization or escaping of the storyboard text is performed before it is interpolated into instructions for sub-agents.
[COMMAND_EXECUTION]: The skill utilizes multiple shell commands to manage project state, initialize repositories, and perform quality checks.
Executes npx hyperframes for project initialization (init), authentication status, usage tracking, linting, and final project verification.
Executes local JavaScript scripts using node to record preferences, apply recipes, and generate worker packets.
[EXTERNAL_DOWNLOADS]: The skill depends on external resources and package runners for tool execution and updates.
Performs skill updates via npx hyperframes skills update, which fetches and installs updated skill logic from a remote registry.
Queries a hosted registry for visual treatments and components using npx hyperframes catalog.

general-video