multi-agent-blueprint_skill

Overview

This skill is a production-tested blueprint for building 5–10 specialized AI agents in OpenClaw with cross-agent routing, Telegram multi-bot support, and role-based architecture. It supplies templates, patterns, and configuration examples to run coordinated agent teams with model tiering, centralized file and DB masters, and cost-saving defaults. Use it to deploy resilient multi-agent systems for business workflows, notifications, and collaborative group sessions.

How this skill works

The blueprint provides agent directory layouts and four required agent files (IDENTITY.md, SOUL.md, AGENTS.md, USER.md) that define role, personality, and routing. Agents communicate via sessions_send using session keys, enabling a Coordinator or direct specialist flows. It includes Telegram per-agent account settings, model fallback chains, memory/RAG setup, and patterns for File Master and Notion Master to centralize sensitive operations.

When to use it

• You need a 3–10 agent team with clear role separation and routing.
• You want Telegram multi-bot DMs plus controlled group @mention behavior.
• You need cost-optimized model tiering and fallback chains for reliability.
• You must centralize file or database access through a single agent (File/DB Master).
• You want persistent memory, RAG, and session compaction without losing key context.

Best practices

• Assign one Coordinator to route requests or let users DM specialists directly for speed.
• Use model tiers: Opus for complex reasoning, Sonnet for standard tasks, Haiku for cheap fast tasks.
• Enable agent-to-agent routing and dmScope: per-account-channel-peer to avoid session collisions.
• Route all filesystem and DB operations through dedicated File Master and Notion/DB Master agents.
• Enable heartbeat for local free models and prompt/memory caching to reduce API costs.

Example use cases

• Small business automation: Coordinator + Tech + Finance + Sales to handle invoices, hosting, and outreach.
• Group brainstorming: Multiple agents in one Telegram group with requireMention enabled for structured sessions.
• RAG-enabled knowledge base: Data agent as Notion master stores and retrieves persistent facts for specialists.
• File workflows: Tech agent as File Master handles uploads, backups, and NAS access via sessions_send.