Repository inventory

spatie/ray-skills

Skills indexed from this repository, with install-style signals scoped to the repo.
1 skills3 GitHub stars0 weekly installsGitHubOwner profile

Overview

This skill teaches how to interact with the Ray application through an LLM-equipped agent. It explains available payload formats, HTTP API usage, and action types so you can send structured content and commands to Ray from code. Use it to reliably format agent outputs for rich rendering inside the Ray UI.

How this skill works

The skill inspects the desired output type and maps it to Ray-compatible payloads (log, table, json, text, html, xml, carbon, custom, decoded-json, file-contents). It also supports agent-focused payloads for rich HTML, markdown, and mermaid rendering. When configured, the skill can propose using a Ray MCP server for delivery and suggests the skill as a secondary option alongside in-chat responses. It exposes action triggers (confetti, show/hide app, clear logs) to control the Ray interface programmatically.

When to use it

  • Whenever the user asks to send content or commands to Ray
  • When you need rich rendering (HTML, Markdown, Mermaid) inside Ray
  • To deliver structured data like tables, JSON, or decoded JSON objects
  • To log runtime events, file contents, or formatted text with preserved whitespace
  • When you want the agent to trigger UI actions in Ray (show/hide, confetti, clear)

Best practices

  • Read the specific rule file for the payload you intend to use to ensure correct fields and examples
  • Prefer the Ray MCP server when available and offer it as the default delivery method if configured
  • Use agent-focused payloads for complex visualizations or diagrams to leverage Ray’s rendering
  • Keep payloads minimal and well-structured: use decoded-json for parsed objects and json for raw strings
  • Use action types sparingly and only when they improve user experience (e.g., confetti for success, clear before new sessions)

Example use cases

  • Send a structured JSON result from a debug run so Ray renders it with syntax highlighting
  • Log key-value diagnostics as a table payload for quick inspection during CI or local debugging
  • Render a generated flowchart using the mermaid agent payload for architecture discussions
  • Push file contents or code snippets with preserved formatting into Ray for review
  • Trigger Ray to show the app and celebrate a completed task with a confetti action

FAQ

If the MCP server is available, prefer it for reliable delivery and recommend it to users; otherwise fall back to direct HTTP requests as described in the rules.

Which payload should I use for parsed JSON objects?

Use the decoded-json payload for parsed JSON objects so Ray can render them as structured data rather than raw strings.

Can the agent control Ray’s UI?

Yes. Use action payloads (show-app, hide-app, confetti, clear-all) to control visibility and simple UI behaviors programmatically.

1 skills

More from this maintainer
Other repositories and skills published under the same GitHub owner.
Skills library
Jump back to the full directory or explore grouped topics.
Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational