🤖 AI Agent Builder Plugin v1.1

Draft idea → Running production agent on your chosen platform.

This plugin gives Claude the ability to guide you through building, setting up, testing, and deploying production-quality AI agents — with actual commands you can run right now, not just code that sits in a file.

---

The Full Pipeline

💡 Draft Idea
    ↓
/analyze → Which AI for which task? (INTELLIGENCE LAYER) ← NEW
    ↓
/spec    → PRD + Architecture Spec (DEFINE)
    ↓
/plan    → Phased Build Roadmap (PLAN)
    ↓
/build   → Production Code / Workflow (BUILD)
    ↓
/setup   → Install & Run Locally RIGHT NOW (SETUP)
    ↓
/test    → Verified & Tested (VERIFY)
    ↓
/review  → Security + Cost QA Gate (REVIEW)
    ↓
/ship    → Production Deployment (SHIP)
✅ Running Agent

---

Commands

Command	Phase	What You Get
/analyze	INTELLIGENCE	Feature-by-feature AI decision table, tool comparison, automation vs human, build order
/spec	DEFINE	Full PRD + architecture spec, platform selection
/plan	PLAN	Phased build roadmap, topology diagram, tech decisions
/build	BUILD	Complete production code / n8n workflow / config files
/setup	SETUP	Copy-paste commands to run the agent on your machine now
/test	VERIFY	Unit tests, behavior tests, integration tests, debug guide
/review	REVIEW	Security audit, cost estimate, reliability check
/ship	SHIP	Production deployment (Docker, Railway, n8n Cloud, VPS)

Or just describe your idea → Claude runs the full pipeline automatically.

---

Supported Platforms

Platform	How to Run	Best For
🟧 OpenClaw	Local gateway, npm install	Personal agent on phone (WhatsApp/Telegram/Slack)
🟪 Hermes Agent	Docker, self-hosted	Autonomous agent with memory + 61 tools
🟩 n8n	Docker local or n8n.io cloud	Visual workflows, 500+ integrations
🟦 Claude Agent SDK	Python/TypeScript, local	Code-first, full developer control
🔵 Claude Cowork	Inside Claude desktop app	File automation, MCP integrations
🤖 Claude Managed Agents	Anthropic-hosted	Zero infra, fully managed

---

What's Included

8 Skills

• agent-architect — Master skill: full pipeline, platform selection, "Run It Now" delivery
• spec — DEFINE: interview + PRD writing with platform selection
• plan — PLAN: architecture diagram, phased tasks, tech decisions
• build — BUILD: production code with typed models, error handling, logging
• setup — SETUP: exact local setup commands per platform ← KEY SKILL
• test — VERIFY: unit/integration/behavior/regression tests
• review — REVIEW: security, cost, reliability QA gate
• ship — SHIP: Docker, Railway, n8n Cloud, VPS deployment

Build References (bundled)

• claude-agent-sdk.md — Python agent patterns, async, orchestrator, memory, retry
• n8n-agent.md — n8n Docker setup, workflow JSON, multi-agent, HITL
• multi-agent.md — Orchestrator+workers, pipeline, critic-actor, context passing
• openclaw-setup.md — OpenClaw local gateway, config, channels, VPS deploy
• hermes-agent-setup.md — Hermes Docker setup, config, 61 tools, custom skills
• cowork-workflow.md — Cowork skill building, MCP connectors, plugin packaging
• patterns.md — ReAct, Plan-Execute, Reflexion, Map-Reduce, HITL patterns

---

Example Usage

Full pipeline (just describe your idea):

"I want an agent that monitors my GitHub issues and automatically drafts replies using my company's FAQ. Runs on my VPS, accessible via Telegram."

Claude will: pick OpenClaw or Hermes → spec → plan → build → give you exact Docker/npm commands to run it.

Individual commands:

/spec    I want a Slack bot that summarizes daily standups
/plan    [after spec is done]
/build   Start Phase 1
/setup   How do I run this locally right now?
/test    Write tests for this
/review  Check before I ship
/ship    Deploy to Railway

---

Version History

• v1.1 (May 2026): Added /setup skill, OpenClaw + Hermes Agent support, "Run It Now" delivery, Cowork workflow reference
• v1.0 (May 2026): Initial release — 7 skills, DEFINE→SHIP pipeline, Claude SDK + n8n + multi-agent

Author: Tanim