The knowledge base for building skills and extensions with effectorHQ.
| Guide | For who | What you'll learn |
|---|---|---|
| Getting Started | New builders | Create your first skill in 10 minutes |
| Architecture | Anyone curious | How effectorHQ fits together: manifests, core, targets, bridges |
| Skill Development | Skill builders | Build language-agnostic skills with SKILL.md |
| Extension Development | TypeScript developers | Build full-featured plugins with TypeScript |
| SKILL.md Format Reference | Everyone | Complete reference for SKILL.md files |
| Manifest Reference | Extension builders | Reference for effector.toml runtime bindings |
Skills: Markdown-first capabilities that live next to code (or under a skills/ tree). The toolchain parses, validates, and compiles them—no single vendor runtime required.
Extensions/Plugins: TypeScript packages with full IDE and deployment stories: channels, config, and heavier logic—published like any other npm module.
Registry: Optional. Manifests stay transport-agnostic; a registry is just one way to distribute them—not part of the core contract.
Core: One kernel reads SKILL.md and effector.toml, checks schema and types, and emits MCP, OpenAI Agents, LangChain, or JSON IR.
Agent Runtime: Whatever can load the compiled artifact (e.g. an MCP host). effectorHQ does not ship or mandate its own agent loop.
- Architecture — how manifests, core, targets, and bridges relate
- Skill Development — authoring
SKILL.md - Extension Development — TypeScript extensions
- SKILL.md format — every frontmatter field
Corrections and clearer explanations are welcome as pull requests.
Prefer plain sentences, concrete code snippets, and short paragraphs.
effectorHQ · Every effector extends the reach.
This project is currently licensed under the Apache License, Version 2.0 。