/summon - Launch AutoMates Agents

Launch one or more agents in separate terminal windows with their identities, knowledge sections, and startup instructions.

Usage

/summon <agent>              # Launch single agent
/summon <agent1>,<agent2>    # Launch multiple agents
/summon team                 # Launch Planner, Builder, Checker
/summon all                  # Launch all 9 agents

Available Agents

Instructions

When this skill is invoked, execute the summon script:

AUTOMATES_ROOT="$(pwd)"

# Check if we're in the AutoMates project
if [[ ! -d "${AUTOMATES_ROOT}/AgenTeam" ]]; then
    echo "Error: Not in the AutoMates project (no AgenTeam/ folder found)"
    exit 1
fi

bash "${AUTOMATES_ROOT}/.claude/skills/summon/summon.sh" "$ARGUMENTS"

What Happens

1. A new terminal window/tab opens for each agent
2. Claude Code starts with that agent's identity loaded via --append-system-prompt
3. The agent reads their memory, dashboard, and knowledge section
4. Agents share files via Dashboard/Work_Space/

Cross-Platform Support

Notes

• Each agent runs as a separate Claude Code instance
• Agents have persistent memory in AgenTeam/<name>/Memory_Logs/
• Fetcher's memory is in Library/Fetcher/Memory_Logs/
• CLAUDE.md is automatically loaded in every session (shared context)
• Coordinate work through shared files in Dashboard/Work_Space/

Part of AutoMates.AI — Your AI dev team