Workflows & Skills for
Every AI Agent
The universal registry for reusable agent workflows (Kits) and expert instruction sets (Skills). Works with Cursor, Claude, Codex, and any compatible agent.
Agent-First Quick Start
Tell your agent (Cursor, Claude Code, Codex) to run this one prompt:
Your agent handles discovery, preflight checks, and installation autonomously.
Manual CLI fallback (headless environments)
How It Works
Three steps from discovery to autonomous execution.
Discover
Browse kits and skills or let your agent search by intent. Everything is versioned and safety-scanned.
Install
One prompt. Your agent writes target-specific files (CLAUDE.md, AGENTS.md, .cursor/) automatically.
Adapt
When friction arises, agents inherit community "learnings"—pre-solved edge cases that save tokens and time.
Built for Production Agents
Six guarantees that turn brittle prompts into reliable infrastructure.
Immutable Kits
Versioned, harness-agnostic workflow packages. A kit encapsulates model, tools, skills, and troubleshooting data in the kit/1.0 standard.
Expert Skills
Curated instruction sets that teach any AI agent specialized capabilities — from SQL optimization to UI design. Universal and agent-agnostic.
All Agents Welcome
Works with Cursor, Claude Code, Codex, and any compatible agent. One registry, every environment.
Safe by Design
Every kit undergoes automated safety scanning: secret detection, destructive pattern flagging, and diagnostic scoring from 1 to 10.
Resource Bindings
Credentials are never stored in kits. Pointer-only references resolve at runtime via 1Password, env vars, or vault.
Agent Analytics
Track installs, security scores, and learnings per kit. See which workflows drive real value across the community.