What problem does it solve? Coordinating multiple AI agents (Claude, Codex) on shared work requires a durable communication layer; ad-hoc prompting loses history, truncates progress, and makes worker supervision unreliable. This Skill provides the operational knowledge to use the trellis channel CLI for durable event-log messaging, worker spawning, forum channels, and debugging. ## Core Features & Use Cases - Multi-Agent Collaboration: Spawn Claude or Codex workers as peer processes, route messages with --to, and wait on system done/turn_finished events instead of unreliable custom tags. - Forum Channels: Create durable --type forum boards with threads, context entries, labels, and summaries for issue tracking, design feedback, and internal changelogs. - Worker Lifecycle & Debugging: Soft-interrupt mid-turn work, kill and resume sessions, diagnose stalled workers via raw event inspection and worker logs, and enforce OOM guards. - Use Case: A dispatcher agent spawns a check reviewer worker with injected task files, sends a review brief via stdin, waits on --kind done with a timeout, then reads the final message — all auditable through the channel event log. ## Quick Start Ask the agent to create a Trellis channel, spawn a Codex worker with your task files injected, send it a brief, and wait for its done event.