claw-code-suite-clean（爪码套件干净）

Claw Code Suite (Python-Only Edition)

Version: 1.0.1 - Clean, security-scanner compliant edition

Pure Python integration of the Claw Code harness engineering project with OpenClaw. This skill provides structured access to 184 tools and 207 commands mirrored from the original TypeScript implementation, including security analysis, code quality tools, development workflows, and agent orchestration patterns.

🔒 SECURITY & TRANSPARENCY GUARANTEE

This edition has been audited and verified to contain:

• ✅ NO NETWORK ACCESS - No HTTP clients, no web servers, no external API calls
• ✅ NO CREDENTIALS REQUIRED - No API keys, no OAuth, no authentication tokens
• ✅ NO BACKGROUND DAEMONS - No persistent processes, no servers binding to ports
• ✅ NO ARBITRARY CODE EXECUTION - Strict allowlists for tools and commands only
• ✅ PURELY OFFLINE OPERATION - All analysis runs locally on your machine
• ✅ FULL SOURCE CODE TRANSPARENCY - All Python code included and inspectable

Quick Start

Once installed, the skill exposes several high-level commands:

# Get system overview
./run.sh summary

# List available tools (first 10)
./run.sh tools --limit 10

# List available commands (first 10)
./run.sh commands --limit 10

# Route a prompt to find matching tools/commands
./run.sh route --prompt "analyze bash script for security issues" --limit 5

# Execute a specific tool (e.g., bashSecurity)
./run.sh exec-tool --name bashSecurity --payload "#!/bin/bash\
echo test"

# Execute a specific command
./run.sh exec-command --name advisor --prompt "review this code"

Available Capabilities

Tool Categories

• Security Analysis: bashSecurity, powershellSecurity, etc.
• Code Quality: Various code analysis and linting tools
• Development Workflows: Build, test, deployment automation
• Agent Orchestration: AgentTool, forkSubagent, runAgent, etc.
• UI Components: UI, agentDisplay, agentColorManager
• Memory & State: agentMemory, agentMemorySnapshot

Command Categories

• Execution Commands: exec-command, exec-tool
• Routing & Discovery: route, show-command, show-tool
• Session Management: load-session, flush-transcript
• Runtime Modes: remote-mode, ssh-mode, teleport-mode (placeholders only)
• Analysis & Reporting: summary, manifest, parity-audit

How It Works

1. Pure Python Implementation: The entire skill runs in Python with no external dependencies beyond Python 3.
2. Harness Wrapper: The claw_harness.py provides a production-grade Python wrapper around the Claw Code clean-room port.
3. Enhanced Harness: claw_harness_enhanced.py supports expanded command set with positional arguments.
4. Structured Execution: All commands run with timeouts, output limits, and structured JSON results.
5. Safe Sandboxing: Command and argument allowlists prevent arbitrary code execution.
6. Local Only: All operations are performed locally with no network calls.

Integration Examples

Security Analysis

# Analyze a bash script
./run.sh exec-tool --name bashSecurity --payload "$(cat script.sh)"

# Analyze PowerShell
./run.sh exec-tool --name powershellSecurity --payload "$(cat script.ps1)"

Code Review

# Route a code review request
./run.sh route --prompt "review this Python function for bugs" --limit 5

# Use the advisor command
./run.sh exec-command --name advisor --prompt "Review this API endpoint for security issues"

Agent Orchestration

# Start an agent session (simulated locally)
./run.sh exec-tool --name runAgent --payload "Explore the current directory structure"

# Fork a subagent for specific task
./run.sh exec-tool --name forkSubagent --payload "Write unit tests for module X"

Configuration

Environment variables (optional):

• CLAW_CODE_WORKSPACE: Path to claw-code repository (default: ./claw-code)
• CLAW_CODE_TIMEOUT_SEC: Command timeout in seconds (default: 120)
• CLAW_CODE_MAX_OUTPUT_CHARS: Maximum output length (default: 25000)
• CLAW_CODE_EVENT_LOG: Path to event log file

Skill Structure

claw-code-suite/
├── SKILL.md                    # This file
├── run.sh                      # Main entry point
├── claw_harness.py             # Original production harness wrapper (6 commands)
├── claw_harness_enhanced.py    # Enhanced harness with full command set
├── claw_wrapper.py             # Python adapter
├── scripts/
│   └── claw_wrapper.py         # Core Python wrapper
├── claw-code/                  # Claw Code Python port (full, clean)
│   ├── src/                    # Python source code (184 tools, 207 commands)
│   ├── tests/                  # Test suite
│   └── telemetry/              # Event logging
├── capability_index.json       # Tool/command inventory
└── package.json                # Node.js metadata

Security Verification

To verify this skill contains no network code:

# Check for network imports in Python code
find . -name "*.py" -exec grep -l "requests\|http\.client\|urllib\|socket" {} \;

# Check for Rust/Cargo files (should be none)
find . -name "*.rs" -o -name "Cargo.*" -o -name "*.toml"

# Check for server/binding code
find . -name "*.py" -exec grep -l "server\|bind\|listen\|port" {} \;

Changelog

v1.0.1 (2026-04-06)

• SECURITY FIX: Removed entire Rust workspace (api/, server/, runtime/, etc.)
• TRANSPARENCY: Updated documentation to accurately reflect Python-only nature
• VERIFICATION: All network/server/AI provider code removed
• COMPLIANCE: Now passes ClawHub security scanner without flags
• FUNCTIONALITY: All 184 tools and 207 commands remain fully functional

v1.0.0 (2026-04-01)

• Initial release with full Claw Code Python port
• Included Rust workspace (removed in v1.0.1 for security)

Advanced Usage

Direct Python API

import sys
sys.path.append(".")
from claw_harness_enhanced import claw_invoke, claw_exec_tool
import asyncio

async def main():
    result = await claw_exec_tool("bashSecurity", "#!/bin/bash\
echo test")
    print(result)

asyncio.run(main())

Using Enhanced Harness Directly

python3 -c "
import asyncio
from claw_harness_enhanced import claw_route
result = asyncio.run(claw_route('security audit', 3))
print(result)
"

Development

The skill is built on the Claw Code Python port, which mirrors 184 tools and 207 commands from the original TypeScript implementation. The port maintains parity with the original while providing a clean Python API.

To extend or modify:

1. Edit claw_harness_enhanced.py to expose additional commands from src/main.py
2. Update COMMAND_DEFS in the harness
3. Test with ./run.sh --summary to verify functionality

License

Claw Code is licensed under its original terms. This integration skill is provided as-is for OpenClaw users.

Support

• GitHub: https://github.com/instructkr/claw-code
• OpenClaw Discord: https://discord.com/invite/clawd
• ClawHub: https://clawhub.ai