What problem does it solve?
This Skill solves the problem of reliably coordinating multiple AI agents to execute a larger software task without context overflow, race conditions, or silently accepted conflicting results.
Core Features & Use Cases
- Topology selection & orchestration patterns: Choose hub-and-spoke, pipeline, peer network, or hierarchical decomposition for the right workflow shape.
- Plan-approval gates: Require the orchestrator to approve a worker sub-plan before execution to prevent drift and scope creep.
- Shared task ownership with file-locking: Use a task registry with optimistic locking to ensure exactly one owner per task and prevent concurrent file writes.
- Dynamic agent spawning with bounded context: Spawn specialist sub-agents when complexity is high, passing a bounded context slice and requiring structured results.
- Conflict resolution & observability: Detect conflicts, escalate or vote per strategy, and log events to support debugging and stall recovery.
- ActiveGraph substrate (event-sourced world state): Replace context-passing with an event-driven graph and enforce status transition rules across task/file/test/patch/review objects.
Quick Start
Use multi-agent-coordination to run a parallel plan for implementing a feature by assigning coding, testing, and review work to distinct agents with locked shared artifacts.
