---
name: multica
description: Use this skill when the user wants to assign coding tasks to AI agents via Multica. Manage issues, agents, runtimes, and daemon through the `multica` CLI. Supports creating issues, assigning to agents, tracking task status, and managing the agent runtime daemon.
metadata:
  builtin_skill_version: "1.0"
  qwenpaw:
    emoji: "🤖"
---

# Multica — Managed Coding Agents

## When to Use

Use this skill when you need to **assign coding tasks to AI agents** managed by Multica, or **check the status** of agent-executed tasks.

### Should Use
- User asks to "assign a task to an agent", "let an AI agent write code", "create a coding task"
- User wants to check if an agent has finished its work
- User mentions Multica, coding agents, or agent task management
- User wants to manage the Multica daemon (start/stop/status)

### Should Not Use
- The task is about general QwenPaw operations (use other skills)
- The task is a simple one-off command (just use shell directly)
- The user wants to schedule recurring tasks (use the `cron` skill)

---

## Prerequisites

1. **Multica CLI** must be installed: `brew install multica-ai/tap/multica` or use the install script
2. **Server connection** must be configured: `multica setup` (cloud or self-host)
3. **Daemon** must be running: `multica daemon start`
4. At least one **Agent** must be created in the workspace
5. At least one **Runtime** must be connected

---

## Quick Setup (First Time)

```bash
# 1. Setup (connects to Multica Cloud or self-host)
multica setup

# 2. Start the daemon
multica daemon start

# 3. Verify runtime is connected
multica runtime list --output json

# 4. Create an agent (via Web UI: Settings → Agents → New Agent)
#    Or via CLI if supported
```

---

## Core Operations

### Daemon Management

```bash
# Check daemon status
multica daemon status

# Start daemon (runs in background)
multica daemon start

# Start daemon in foreground (for debugging)
multica daemon start --foreground

# Stop daemon
multica daemon stop

# Restart daemon
multica daemon restart

# View daemon logs
multica daemon logs
```

### Agent Management

```bash
# List all agents
multica agent list --output json

# Get agent details
multica agent get <agent_id> --output json

# Create an agent (usually done via Web UI)
multica agent create --name "coder-1" --runtime <runtime_id> --provider claude-code

# Update agent
multica agent update <agent_id> --name "new-name"

# Archive an agent
multica agent archive <agent_id>

# List agent skills
multica agent skills <agent_id>
```

### Issue Management (Task Assignment)

```bash
# List all issues
multica issue list --output json

# List issues with status filter
multica issue list --status open --output json

# Create a new issue (task) for an agent
multica issue create \
  --title "Implement user authentication" \
  --description "Add JWT-based auth to the API" \
  --assignee <agent_id> \
  --repo <repo_id> \
  --output json

# Assign an existing issue to an agent
multica issue assign <issue_id> --agent <agent_id>

# Get issue details (including status, assignee, runs)
multica issue get <issue_id> --output json

# Check issue status
multica issue status <issue_id>

# Change issue status
multica issue status <issue_id> --status resolved

# Add a comment to an issue
multica issue comment <issue_id> --body "Task completed, please review"

# List execution runs for an issue
multica issue runs <issue_id> --output json

# List messages from an agent's execution
multica issue run-messages <issue_id> --output json

# Cancel a running task
multica issue cancel-task <issue_id>

# Rerun an issue (fresh task)
multica issue rerun <issue_id>

# Search issues
multica issue search "authentication" --output json

# Show token usage for an issue
multica issue usage <issue_id>
```

### Runtime Management

```bash
# List runtimes
multica runtime list --output json

# Check runtime activity
multica runtime activity <runtime_id>

# Get token usage
multica runtime usage <runtime_id>
```

### Squad Management

```bash
# List squads
multica squad list --output json

# Assign issue to a squad (leader routes to best agent)
multica issue create \
  --title "Fix login bug" \
  --assignee <squad_id> \
  --output json
```

---

## Workflow: Assign Task to Agent

```
1. Check daemon is running:     multica daemon status
2. List available agents:       multica agent list --output json
3. List available repos:        multica repo list --output json
4. Create issue + assign:       multica issue create --title "..." --description "..." --assignee <agent_id> --repo <repo_id>
5. Monitor progress:            multica issue get <issue_id> --output json
6. Check execution messages:    multica issue run-messages <issue_id> --output json
7. Review result:               multica issue get <issue_id> --output json
```

## Workflow: Check Task Status

```
1. List issues:                 multica issue list --output json
2. Get specific issue:          multica issue get <issue_id> --output json
3. View execution logs:         multica issue run-messages <issue_id> --output json
```

## Workflow: Cancel / Rerun Task

```
1. Cancel running task:         multica issue cancel-task <issue_id>
2. Rerun as fresh task:         multica issue rerun <issue_id>
```

---

## Output Formats

All list/get commands support `--output json` for machine-readable output. Use this when parsing results programmatically.

---

## Error Handling

| Error | Solution |
|-------|----------|
| `No server configured` | Run `multica setup` to connect to Multica Cloud or self-host |
| `Daemon: stopped` | Run `multica daemon start` |
| `No agents found` | Create an agent via Web UI or `multica agent create` |
| `No runtimes found` | Ensure daemon is running and connected |
| `Workspace not found` | Set workspace: `multica config set workspace-id <id>` |

---

## Integration with QwenPaw

This skill enables QwenPaw to act as a **task dispatcher** for Multica's coding agents:

1. **7爷** tells QwenPaw what needs to be done
2. QwenPaw breaks down the task and decides if it needs a coding agent
3. If yes, QwenPaw uses this skill to create a Multica issue and assign it
4. QwenPaw monitors progress and reports back to 7爷
5. When done, QwenPaw can review results and provide feedback

This makes Multica agents **sub-agents** of QwenPaw's orchestration.

---

## Help

```bash
multica --help
multica issue --help
multica agent --help
multica daemon --help
multica runtime --help
multica squad --help
```