multi-agent-patterns-advanced
Installation
SKILL.md
Multi-Agent Patterns — Advanced
See the core
multi-agent-patternsskill for basics: orchestration vs choreography, intent classification routing, in-memory state, handoffs, fan-out/fan-in, Claude SDK loop, and observability.
When to Activate
- Building a capability registry for dynamic agent dispatch
- Implementing durable cross-agent state (Redis, DynamoDB event log)
- Decomposing complex goals into parallel and sequential sub-tasks
- Writing deterministic tests for multi-agent workflows
- Selecting the right pattern from the quick-selection guide
- Managing failure modes across agent boundaries
- Controlling token cost in large multi-agent systems
- Isolating parallel file-editing agents with git worktrees