skill-porter

name	skill-porter
description	Convert skills from ClawHub, Claude, Gemini, Codex formats to Nanobot-native format. Use when user wants to port/migrate/convert skills to work with Nanobot (HKUDS/nanobot). Primary focus: ClawHub skills. Keywords: convert skill, port skill, migrate skill, nanobot skill, clawhub to nanobot.

Skill Porter

Convert skills from various AI agent ecosystems to Nanobot-native format.

Supported Sources

Source	Priority	Status
ClawHub	Primary	Full support
Claude Skills	Secondary	Full support
Gemini Skills	Secondary	Full support
Codex Skills	Secondary	Full support

Nanobot Architecture Reference

Nanobot is a Python-based ultra-lightweight AI agent. Key characteristics:

Memory Structure

workspace/
├── SOUL.md           # Bot personality (Dream-managed, DO NOT EDIT)
├── USER.md           # User profile (Dream-managed, DO NOT EDIT)
├── memory/
│   ├── MEMORY.md     # Long-term facts (Dream-managed)
│   └── history.jsonl # Append-only JSONL history

Built-in Tools

• read_file, write_file, edit_file - File operations
• list_directory - Directory listing
• exec - Shell command execution
• grep - Search files (use for history.jsonl)
• web_search, web_fetch - Web operations

MCP Tools

Configure in ~/.nanobot/config.json:

{
  "tools": {
    "mcpServers": {
      "server-name": {
        "command": "npx",
        "args": ["-y", "@scope/server-package"],
        "enabledTools": ["tool1", "tool2"]
      }
    }
  }
}

Conversion Process

Step 1: Analyze Source Skill

Use read_file to read the source SKILL.md and identify:

1. Skill type: Prompt-only vs tool-dependent
2. Tool dependencies: Which tools/commands are referenced
3. Path references: Hardcoded paths that need adjustment
4. Source-specific features: Plugins, commands, lifecycle hooks

Step 2: Determine Compatibility

High compatibility (direct port):

• Pure prompt/rule skills
• No source-specific tool names
• No hardcoded source paths

Medium compatibility (needs adaptation):

• Uses tool names that differ between systems
• References source config paths
• Uses source-specific commands

Low compatibility (needs rewrite):

• Strong plugin dependencies
• Lifecycle hooks
• Marketplace features

Step 3: Apply Transformations

Tool Name Mappings

Source Tool	Nanobot Equivalent
read	read_file
write	write_file
edit	edit_file
exec	exec (same)
browser	MCP: Playwright
message	Channel config
cron	Nanobot cron
memory_search	grep on history.jsonl

Path Mappings

Source Path	Nanobot Path
~/.openclaw/	~/.nanobot/
{workspace_root_dir}	{workspace}
{bundled_skill_dir}	nanobot/skills/
MEMORY.md	memory/MEMORY.md
memory/YYYY-MM-DD.md	memory/history.jsonl

Command Mappings

Source Command	Nanobot Command
openclaw gateway	nanobot gateway
/status	/status (same)

Step 4: Generate Output

Create the converted SKILL.md with:

1. Preserved frontmatter (name, description)
2. Transformed tool references
3. Updated path references
4. MCP requirements if needed

Step 5: Document Migration

Create MIGRATION.md with:

• What was changed
• What needs manual setup
• What couldn't be converted

Scripts

scripts/convert_skill.py

Main conversion script.

python scripts/convert_skill.py <source_path> [--output <output_path>] [--dry-run] [--verbose]

Options:

• --output: Output directory
• --dry-run: Show changes without writing
• --verbose: Detailed output

scripts/batch_convert.py

Convert multiple skills.

python scripts/batch_convert.py <input_dir> --output <output_dir>

scripts/validate_nanobot.py

Validate a skill for Nanobot compatibility.

python scripts/validate_nanobot.py <skill_path>

Example Usage

Convert a single ClawHub skill:

Convert the weather-advisor skill from ClawHub to Nanobot format.

Batch convert:

Convert all skills in ~/my-skills/ to Nanobot format.

Check compatibility:

Check if the email-skill is compatible with Nanobot.

Important Notes

1. Dream-managed files: SOUL.md, USER.md, MEMORY.md are auto-managed. Do NOT include instructions to edit these files.

2. MCP for missing tools: If a skill needs tools Nanobot doesn't have built-in, generate MCP server config.

3. Channel differences: Nanobot uses channel-specific configs (Telegram, Discord, etc.) instead of a unified message tool.

4. Memory access: Use grep tool to search memory/history.jsonl. Format: JSONL with cursor, timestamp, content fields.

5. Config location: Nanobot config is at ~/.nanobot/config.json.

Limitations

These features have no Nanobot equivalent:

• browser tool (use MCP Playwright)
• canvas tool (no equivalent)
• nodes tool (no equivalent)

Output Format

Converted skills are output as:

<skill-name>-nanobot/
├── SKILL.md              # Converted skill
├── MIGRATION.md          # Migration notes
└── mcp_requirements.json # MCP servers needed (if any)