name: Slack Agents description: Manage pi-mono Slack agents (mom bots). Activate when user mentions slack agent, pilot, pilot sandbox, zip agent, revive agent, start agent, stop agent, agent status, pi-mono agents, or needs to manage running Slack bot instances.

Slack Agents Management

Multi-agent Slack bot architecture based on pi-mono (Mario Zechner's mom package).

Architecture Overview

pi-mono/                          # Primary fork (Pilot agent)
└── packages/mom/
    ├── run.sh                    # Direct start script
    ├── start-pilot.sh            # Tmux launcher
    ├── .env                      # Slack tokens + Anthropic key
    ├── data/                     # Workspace (mounted to container)
    └── logs/                     # Daily log files

pi-mono-zip/                      # Worktree of pi-mono (Zip agent)
└── packages/mom/
    ├── run.sh
    ├── start-zip.sh              # Tmux launcher
    ├── .env
    ├── data/
    └── logs/

Note: pi-mono-zip is a git worktree of pi-mono, not a separate clone.

Current Agents

Common Operations

Check Agent Status

# List all agent containers
docker ps --format 'table {{.Names}}\t{{.Status}}' | grep -E 'sandbox|mom'

# List tmux sessions
tmux list-sessions | grep -E 'pilot|zip'

# Check if bot process is running
pgrep -af "mom/dist/main.js"

Start an Agent (preferred: tmux)

# Pilot (in tmux session)
cd ~/code/pi-mono/packages/mom && ./start-pilot.sh

# Zip (in tmux session)
cd ~/code/pi-mono-zip/packages/mom && ./start-zip.sh

# Attach to running session
tmux attach -t pilot
tmux attach -t zip

Start an Agent (direct, no tmux)

cd ~/code/pi-mono/packages/mom && ./run.sh
cd ~/code/pi-mono-zip/packages/mom && ./run.sh

Stop an Agent

# Kill tmux session (also kills the process)
tmux kill-session -t pilot
tmux kill-session -t zip

# Or kill the node process directly
pkill -f "pi-mono/packages/mom/dist/main.js"      # Pilot
pkill -f "pi-mono-zip/packages/mom/dist/main.js"  # Zip

View Logs

# Today's log
tail -f ~/code/pi-mono/packages/mom/logs/mom-$(date +%Y-%m-%d).log

# Recent activity
tail -100 ~/code/pi-mono-zip/packages/mom/logs/mom-$(date +%Y-%m-%d).log

# Or attach to tmux session to see live output
tmux attach -t pilot

Restart Container (if needed)

docker restart pilot-sandbox
docker restart zip-sandbox

Environment Setup

Each agent needs .env with:

MOM_SLACK_APP_TOKEN=xapp-...    # Socket mode token
MOM_SLACK_BOT_TOKEN=xoxb-...    # Bot OAuth token
ANTHROPIC_API_KEY=sk-ant-...    # Or ANTHROPIC_OAUTH_TOKEN

Adding a New Agent

1. Create worktree (preferred) or clone:

   cd ~/code/pi-mono
   git worktree add ../pi-mono-newagent -b newagent-deploy
   

2. Configure container name in run.sh:

   CONTAINER_NAME="newagent-sandbox"
   

3. Create tmux launcher (copy from existing):

   cp start-pilot.sh ../pi-mono-newagent/packages/mom/start-newagent.sh
   # Edit SESSION="newagent"
   

4. Create Slack app:

   • https://api.slack.com/apps
   • Enable Socket Mode
   • Add bot scopes (see packages/mom/docs/slack-bot-minimal-guide.md)
   • Generate tokens

5. Configure .env:

   cd ~/code/pi-mono-newagent/packages/mom
   cp .env.example .env
   # Edit with tokens
   

6. Build and run:

   cd ~/code/pi-mono-newagent
   npm install
   npm run build
   cd packages/mom
   ./start-newagent.sh
   

Troubleshooting

Agent not responding

1. Check container: docker ps | grep sandbox
2. Check tmux: tmux list-sessions
3. Check logs for errors
4. Restart: tmux kill-session -t pilot && ./start-pilot.sh

"Container not found"

# Recreate container (data preserved in ./data/)
docker rm pilot-sandbox 2>/dev/null
./run.sh  # Will recreate

Token errors

• Verify tokens in .env
• Check Slack app is installed to workspace
• Ensure Socket Mode is enabled

Future Direction

This architecture enables:

• Multiple isolated agents per workspace
• Shared codebase with per-agent config (worktrees)
• Independent sandboxed execution environments
• Named tmux sessions for easy management
• Centralized management from TinyBat VPS