name: worktree-operations description: Guide for working within git worktrees in the Orient monorepo. Use this skill when you detect you're working in a worktree (path contains "skill-worktrees", "app-worktrees", or is outside the main project directory), when asked to "set up worktree", "install dependencies in worktree", "build in worktree", "merge worktree", or when troubleshooting worktree-specific issues. Covers pnpm/turbo commands, environment setup, git workflows (merging, conflict resolution, cleanup), and common worktree pitfalls.

Worktree Operations

Detecting a Worktree

You're in a worktree if:

• Path contains skill-worktrees or claude-worktrees (e.g., ~/claude-worktrees/orienter/my-feature-123456)
• Path is outside the main project directory but contains project files
• Branch name starts with skill/ or worktree/

Worktrees are isolated checkouts used for skill development without affecting the main checkout.

Multi-Instance Support

Worktrees automatically get unique instance IDs for running multiple bot instances in parallel:

• Main repo: Instance 0 (original ports)
• Worktrees: Instance 1-9 (auto-assigned, ports offset by instance_id × 1000)

See the multi-instance-development skill for complete details.

Slack Bot Instance Isolation (CRITICAL)

Only ONE Slack bot instance can be connected at a time. Unlike WhatsApp which uses phone-based sessions, Slack uses Socket Mode which means all instances try to connect to the same workspace. Running multiple Slack bots simultaneously causes:

• Connection conflicts and dropped messages
• Duplicate responses
• Session instability

Before Testing Slack Features

Always ensure only one Slack bot instance is running:

# 1. Kill ALL Slack bot processes across all worktrees
pkill -f "tsx.*slack-bot" || true
pkill -f "node.*bot-slack" || true

# 2. Verify no orphaned processes
ps aux | grep -E "slack.*tsx|tsx.*slack" | grep -v grep

# 3. Stop other worktree dev environments
# If you have other worktrees running ./run.sh dev, stop them first:
cd ~/claude-worktrees/other-worktree
./run.sh dev stop

# 4. Start Slack bot in your target worktree only
./run.sh dev

Checking for Running Slack Instances

# Quick check: Count Slack bot processes
ps aux | grep -E "slack.*tsx|tsx.*slack|bot-slack" | grep -v grep | wc -l
# Should be 0 before starting, 1-2 after (tsx watch spawns child process)

# Detailed view: See all Slack-related processes
ps aux | grep -E "slack.*tsx|tsx.*slack|bot-slack" | grep -v grep

# Check which ports are in use (Slack doesn't use a port, but check related services)
lsof -i :4099  # OpenCode server

Killing Orphaned Slack Processes

# If you see multiple Slack processes or stale processes:

# Option 1: Kill by pattern (recommended)
pkill -f "tsx.*slack-bot"
pkill -f "tsx.*watch.*slack"

# Option 2: Kill specific PIDs
ps aux | grep "slack" | grep -v grep | awk '{print $2}' | xargs kill -9

# Option 3: Use run.sh stop (cleans up properly)
./run.sh dev stop

Best Practices for Multi-Worktree Development with Slack

1. Designate one worktree for Slack testing - Don't run Slack in multiple worktrees
2. Stop Slack in main repo when testing in worktree - Run ./run.sh dev stop in main repo first
3. Use --no-slack flag when you don't need Slack: ./run.sh dev --no-slack
4. Check logs for connection issues:

   tail -f logs/instance-0/slack-bot.log | grep -E "(connected|error|conflict)"
   

5. Verify single connection before testing:

   # Should see exactly one "Now connected to Slack" message
   grep "Now connected to Slack" logs/instance-0/slack-bot.log | tail -1
   

Troubleshooting Slack Connection Issues

Symptom: Bot doesn't respond to messages

# Check if bot is connected
grep -E "(connected|error|disconnect)" logs/instance-0/slack-bot.log | tail -10

# Kill all instances and restart fresh
pkill -f "slack-bot"
sleep 2
./run.sh dev

Symptom: Duplicate responses or messages lost

# Multiple bots are likely running - kill all and restart one
pkill -f "tsx.*slack"
./run.sh dev stop  # In ALL worktrees
./run.sh dev       # In ONLY the worktree you want to use

Symptom: "socket hang up" or connection errors

# OpenCode server might be down
curl -s http://localhost:4099/global/health || echo "OpenCode not running"

# Restart everything
./run.sh dev stop && ./run.sh dev

Initial Setup (After Worktree Creation)

When a worktree is first created, run these commands:

# Copy environment from main worktree (already done by cursor worktrees.json)
# cp $ROOT_WORKTREE_PATH/.env .env

# Install dependencies with pnpm (runs automatically via worktrees.json)
pnpm install

Database Setup in Worktrees

Worktrees can use either a shared database (default) or an isolated database (for schema testing).

Shared Database (Default)