The master entry point for all multi-domain engineering tasks. Brainstorms, plans, coordinates specialist agents, and synthesizes results — with full session continuity across conversations.
When to Use
- Tasks that span multiple domains (AI + DevOps, security + data, frontend + backend)
- Complex projects requiring structured plan → execute → verify workflow
- Multi-session projects where state must be preserved across conversations
- Any task where you'd want a senior staff engineer to coordinate the work
Available Domains
| Domain | Lead Agent | Trigger Keywords |
|---|---|---|
| AI/ML | AI/ML Lead | chatbot, LLM, RAG, model, prediction, embeddings, agent |
| Platform | Platform Lead | deploy, kubernetes, CI/CD, infrastructure, cloud, docker |
| Security | Security Lead | PII, compliance, IAM, vulnerability, GDPR, security |
| Data | Data Lead | pipeline, ETL, warehouse, quality, catalog, database |
| Product | Product Lead | feature, UI, API, testing, documentation, frontend, backend |
| Brainstorm | Brainstorm Architect | architecture, solution, design, tradeoff, "best way to" |
Workflow
Resume (check prior session state)
→ Understand (explore project + analyze request)
→ Plan (select agents, sequence work, present for approval)
→ Execute (spawn lead agents in parallel batches)
→ Verify (run tests, check quality gates)
→ Synthesize (report results + save state)
Full Agent Instructions
See AGENTS.md for complete orchestrator protocol.