agent-council

Research agent:

scripts/create-agent.sh \
  --name "Watson" \
  --id "watson" \
  --emoji "🔬" \
  --specialty "Deep research and competitive analysis" \
  --model "skillboss/pilot" \
  --workspace "$HOME/agents/watson" \
  --discord-channel "1234567890"

Image generation agent:

scripts/create-agent.sh \
  --name "Picasso" \
  --id "picasso" \
  --emoji "🎨" \
  --specialty "Image generation and editing specialist" \
  --model "skillboss/pilot" \
  --workspace "$HOME/agents/picasso" \
  --discord-channel "9876543210"

Health tracking agent:

scripts/create-agent.sh \
  --name "Nurse Joy" \
  --id "nurse-joy" \
  --emoji "💊" \
  --specialty "Health tracking and wellness monitoring" \
  --model "skillboss/pilot" \
  --workspace "$HOME/agents/nurse-joy" \
  --discord-channel "5555555555"

Discord Category ID

Option 1: Command line

python3 scripts/setup-channel.py \
  --name channel-name \
  --context "Purpose" \
  --category-id "1234567890"

Option 2: Environment variable

export DISCORD_CATEGORY_ID="1234567890"
python3 scripts/setup-channel.py --name channel-name --context "Purpose"

Finding Discord IDs

Enable Developer Mode:

Copy IDs:

When to use Discord bindings:

When to use sessions_send:

When to use sessions_spawn:

Example: Research Workflow

// Main agent receives request: "Research competitor X"
// 1. Check if Watson is active
const agents = sessions_list({ kinds: ["agent"] })
// 2. Delegate to Watson
sessions_send({
  label: "watson",
  message: "Research competitor X: products, pricing, market position. Write findings to memory/research-X.md"
})
// 3. Watson works independently:
//    - Searches web
//    - Analyzes data
//    - Updates memory file
//    - Reports back when done
// 4. Main agent retrieves results
const results = Read("agents/watson/memory/research-X.md")
// 5. Share with user
"Research complete! Watson found: [summary]"

Communication Flow

Main Agent (You) ↔ Specialized Agents:

User Request
    ↓
Main Agent (Claire)
    ↓
sessions_send("watson", "Research X")
    ↓
Watson Agent
    ↓
Uses web_search
Uses web_fetch
Updates memory files
    ↓
Responds to main session
    ↓
Main Agent synthesizes and replies

Discord-Bound Agents:

User posts in #research channel
    ↓
Watson Agent (bound to channel)
    ↓
Sees message directly
Responds in channel
No main agent involvement

Hybrid Approach:

User: "Research X" (main channel)
    ↓
Main Agent delegates to Watson
    ↓
Watson researches and reports back
    ↓
Main Agent: "Done! Watson found..."
    ↓
User: "Show me more details"
    ↓
Main Agent: "@watson post your full findings in #research"
    ↓
Watson posts detailed report in #research channel