skill-to-agent

# Skill-to-Agent

Transform any OpenClaw skill into a fully functional agent with proper workspace configuration and session management. Solves common failures like `thread=true is unavailable` and `mode="session" requires thread=true`.

## Quick Start

1. **Analyze skill** - Read SKILL.md to understand requirements
2. **Prepare workspace** - Create agent directory in `~/.openclaw/agents/`
3. **Configure agent** - Set up identity, tools, and memory strategy
4. **Spawn with correct binding** - Choose appropriate mode based on channel

## Workflow

### Phase 1: Skill Analysis
- Extract metadata from SKILL.md (name, description, triggers)
- Identify required tools and dependencies
- Determine agent type: isolated, thread-bound, or run-mode

### Phase 2: Workspace Setup
- Create agent directory with proper structure
- Copy skill files (references/, scripts/)
- Configure AGENTS.md, TOOLS.md, USER.md

### Phase 3: Agent Configuration
- Create SOUL.md and IDENTITY.md with skill personality
- Map skill tools to agent capabilities
- Set up memory strategy (MEMORY.md, memory/ directory)

### Phase 4: Agent Registration
- Add agent to OpenClaw's `agents.list` configuration
- Use `gateway` tool with `config.patch` action
- Include agent ID, name, workspace, and agentDir
- Gateway will restart to apply configuration

### Phase 5: Session Spawning
- Choose spawn strategy based on channel capabilities
- Execute with correct binding (thread-bound, isolated, run-mode)
- **CRITICAL:** Always set `cwd` to agent workspace directory
- Verify agent activation and tool access

## Reference Materials

For detailed guidance, refer to:

- **Usage Guide** - [usage-guide.md](references/usage-guide.md) - Complete workflow with examples
- **Troubleshooting** - [troubleshooting.md](references/troubleshooting.md) - Common errors and solutions
- **War Room Example** - [war-room-example.md](references/war-room-example.md) - Detailed conversion example

## Scripts

- `scripts/skill_to_agent.js` - Mai...

# Skill-to-Agent

Transform OpenClaw skills into fully functional, properly configured agents with correct workspace setup, binding strategies, and system registration.

## Overview

This skill provides a methodology for converting any OpenClaw skill into an autonomous agent. It solves common agent creation failures including thread binding errors, workspace isolation issues, file access problems, and improper session spawning.

## Quick Start

### The 5-Phase Methodology

1. **Skill Analysis** - Extract requirements from SKILL.md
2. **Workspace Setup** - Create agent directory with proper structure
3. **Agent Configuration** - Craft identity, memory, and tool access
4. **Agent Registration** - Add to OpenClaw's `agents.list` configuration
5. **Session Spawning** - Launch with correct binding for the channel

### Critical Success Factors

- **Always set `cwd`** - Agent workspace directory for file access
- **Always register agents** - Use `gateway config.patch` to add to configuration
- **Choose correct binding** - Based on channel capabilities (thread-bound, isolated, run-mode)

## Common Problems Solved

### Thread Binding Errors
- `thread=true is unavailable because no channel plugin registered subagent_spawning hooks`
- `mode="session" requires thread=true so the subagent can stay bound to a thread`

### Workspace Issues  
- Agents can't find identity files (IDENTITY.md, SOUL.md)
- File permission and access problems
- Memory configuration failures

### Registration Problems
- Agents not recognized by OpenClaw system
- Configuration not applied after creation

## File Structure

```
skill-to-agent/
├── README.md           # This overview
├── SKILL.md            # Main skill documentation
├── _meta.json          # Skill metadata
└── references/         # Detailed guides
    ├── usage-guide.md      # Complete workflow
    ├── troubleshooting.md  # Error solutions  
    └── war-room-example.md # Conversion example
```

## Usage Examples

### Create a Weather Agent...