Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Add error handling docs, environment variables table, remove hardcoded paths. Progressive disclosure structure.
What This Skill Does
Runs Claude Code as a persistent tmux session, letting you start, send prompts to, and monitor Claude across multiple projects from a single relay script. Session names are deterministic and tied to project paths or aliases. Designed for orchestration and session management, not code reasoning.
Persistent tmux sessions survive terminal disconnects, so Claude keeps working even after you close the window.
When to use it
- Keeping a Claude session alive while switching terminals
- Sending follow-up prompts to a background Claude process
- Monitoring Claude output on a long-running task
- Managing Claude sessions across several codebases at once
- Restarting a hung Claude process without losing project context
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Start a Claude session for my backend project and ask it to review the auth module
- 1Runs `scripts/relay.sh start backend` to create a named tmux session
- 2Resolves 'backend' via projects.map or directory lookup under CLAUDE_RELAY_ROOT
- 3Sends `scripts/relay.sh send backend 'Review the auth module'`
- 4Waits RELAY_WAIT seconds, then runs `scripts/relay.sh tail backend` to read output
- 5Returns Claude's response to the user
Claude's review of the auth module, read from the tmux session buffer
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.