Skip to content

ai-skills.md

Latest commit

 

History

History
63 lines (42 loc) · 2.14 KB

ai-skills.md

File metadata and controls

63 lines (42 loc) · 2.14 KB

AI Skills

Fizz ships an installable agent skill for the core runtime and the React integration.

This skill is optimized for agents, not as a replacement for the normal API docs. Its job is to help coding agents recognize when a task should use Fizz patterns and then apply the library consistently.

Install

Install the skill from skills.sh with:

npx skills add tdreyno/fizz

What The Skill Covers

  • Core @tdreyno/fizz state-machine modeling
  • Runtime setup with createInitialContext(...), createRuntime(...), beforeEnter(...), and enter()
  • Async and scheduling helpers such as startAsync(...), requestJSONAsync(...), timers, intervals, and frame-based work
  • React integration through @tdreyno/fizz-react and useMachine(...)

Scope Boundaries

The published skill intentionally focuses on:

  • @tdreyno/fizz
  • @tdreyno/fizz-react

It intentionally excludes:

  • @tdreyno/fizz-svelte
  • application-specific patterns from react-example/
  • general non-Fizz architecture advice

How Agents Should Use It

Agents should reach for this skill when a task involves:

  • creating or refactoring a Fizz machine
  • reviewing Fizz transitions, actions, or effects
  • wiring a runtime correctly
  • adding async request flows or scheduled behavior
  • writing or reviewing tests for Fizz machines and runtimes
  • integrating an existing machine into React with useMachine(...)

The skill keeps the top-level guidance short and points agents to deeper reference files only when they need more detail.

What Users Should Expect

After the skill is installed, compatible agents can use it automatically when a task matches the Fizz domain. In practice that means the agent should:

  • model workflows with explicit Fizz states and actions
  • preserve the runtime lifecycle Fizz expects
  • use the built-in async and scheduling helpers instead of ad-hoc alternatives
  • prefer controlled runtime drivers for deterministic async and timer tests
  • keep React components thin when a Fizz machine is already the source of truth

Related Docs