This hooks Claude up to Ray, the debugging tool that lets you send logs, tables, JSON, HTML, and other data types to a separate visual window instead of cluttering your terminal. It supports the Ray MCP server or falls back to direct HTTP calls. The agent payload types are notable: you can send rendered Markdown, Mermaid diagrams, or rich HTML straight from Claude's output, which makes it surprisingly good for debugging LLM responses or showing structured results visually. Works across PHP, JavaScript, Python, Ruby, and a dozen other languages. If you're already using Ray for debugging, this turns Claude into something that can actually log to it intelligently instead of just dumping console output.
npx -y skills add spatie/ray-skills --skill ray --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills