Claude Code Openclaw Skill

Integrate OpenClaw with Claude Code CLI to read, edit, run commands, review, and automate workflows on your codebase using AI assistance.

Security Scan

The skill's instructions claim to integrate with the Claude Code CLI, but important operational details (required credentials, install provenance, and token usage) are not declared in the registry metadata and the README endorses running a remote install script, so the package is internally inconsistent and requires caution.

Purpose & Capability

The SKILL.md describes reading/editing the codebase, running shell commands, creating commits/PRs, and configuring MCP servers (which can accept third‑party tokens). The registry metadata declares no required environment variables, credentials, or config paths — yet the instructions explicitly reference ANTHROPIC_API_KEY and examples that accept tokens (e.g., github token for MCP). This mismatch between claimed purpose and declared requirements is unexplained and disproportionate.

Instruction Scope

The runtime instructions instruct users/agents to install and invoke a CLI that will read and modify the repository and run arbitrary shell commands. They also recommend adding external MCP servers with tokens. The SKILL.md endorses executing a remote install script (curl https://claude.ai/install.sh | bash) and contains examples that accept sensitive tokens; these steps broaden the agent's runtime authority beyond what the registry metadata declares.

Install Mechanism

There is no formal install spec in the registry, but the SKILL.md directs users to run a remote install script via curl | bash. Download-and-execute from a remote URL is higher risk unless the URL and release provenance are verified. The registry does not provide a homepage or verified source to confirm the legitimacy of that script.

Credentials

The registry lists no required env vars, but the documentation explicitly references ANTHROPIC_API_KEY and shows examples that embed tokens for MCP (e.g., github_token). Requesting API keys and external tokens is reasonable for a CLI, but not declaring them in metadata is inconsistent and prevents automated permission checks; this raises the possibility of undisclosed credential use or exfiltration.

Persistence & Privilege

The skill is instruction-only, has no install spec in the registry, and does not set always:true. Autonomous invocation is enabled (default) which is normal for skills. The skill does not request permanent platform-level privileges in metadata.

What to consider before installing

This skill may be what it claims (a Claude Code CLI integration), but the README and the registry don't match and the README tells you to run a remote install script. Before installing or invoking it: 1) verify the publisher and homepage (ask for an official source or link to the vendor's documentation), 2) do not run curl | bash blindly — inspect the install script on the remote site or prefer an OS package manager or GitHub release, 3) require the skill author to declare required env vars (e.g., ANTHROPIC_API_KEY) in metadata so you can audit permissions, 4) avoid supplying unrelated credentials (GitHub tokens, other API keys) unless you trust the integration and have scoped tokens with least privilege, 5) run initial tests in an isolated environment (throwaway VM/container) where repository secrets cannot be exfiltrated, and 6) request the author to provide an explicit install spec and homepage so provenance can be validated. If the author provides an official homepage, verified install artifacts (e.g., GitHub release or package registry), and updates the metadata to list required env vars and config paths, my confidence would increase.

Claude Code Skill for OpenClaw

Call Claude Code CLI from OpenClaw to read codebase, edit files, run commands, and automate coding tasks.

Overview

This skill provides integration with Claude Code CLI, allowing OpenClaw to leverage Claude Code's powerful coding capabilities including:

• Reading and understanding codebases
• Editing files with AI assistance
• Running terminal commands
• Creating commits and PRs
• Code review and debugging
• Running sub-agents for parallel tasks

Prerequisites

1. Install Claude Code CLI:

   curl -fsSL https://claude.ai/install.sh | bash
   

2. Authenticate:

   claude auth login
   

3. Verify installation:

   claude --version
   

Usage

Basic Query

Ask Claude Code a question about your codebase:

/claude-code What does the auth module do?

Run a Task

Execute a coding task:

/claude-code write tests for the login function

Continue Previous Session

Resume the most recent conversation:

/claude-code --continue

Resume Specific Session

Resume a session by name or ID:

/claude-code --resume auth-fix

Code Review

Review code changes:

/claude-code review the recent changes

Create Commit

Commit changes with a descriptive message:

/claude-code commit my changes

Run with Custom Model

Use a specific model:

/claude-code --model opus explain this code

JSON Output

Get structured JSON output for scripting:

/claude-code --output-format json --print "list all functions"

CLI Reference

Common Commands

Command	Description
claude -p "query"	Query via SDK, then exit
claude -c	Continue most recent conversation
claude -r "session" "query"	Resume session by ID or name
claude --print "query"	Print response without interactive mode
claude update	Update to latest version
claude auth status	Show authentication status

Useful Flags

Flag	Description	Example
--print, -p	Non-interactive mode	claude -p "query"
--continue, -c	Continue session	claude -c
--resume, -r	Resume specific session	claude -r session-id "task"
--model	Specify model	claude --model opus "task"
--output-format	Output format (text/json/stream-json)	claude -p --output-format json "task"
--max-turns	Limit agentic turns	claude -p --max-turns 3 "task"
--max-budget-usd	Max API spend	claude -p --max-budget-usd 5 "task"
--add-dir	Additional directories	claude --add-dir ../lib "task"
--mcp-config	Load MCP servers	claude --mcp-config ./mcp.json

Environment Variables

Variable	Description
ANTHROPIC_API_KEY	API key for authentication
CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD	Enable CLAUDE.md from additional dirs

Integration with OpenClaw

This skill is designed to work with OpenClaw's exec tool. Here's how it integrates:

Method 1: Direct CLI Execution

Use OpenClaw's exec tool to run Claude Code commands:

claude -p "What files were changed in the last commit?"

Method 2: Session Management

Manage Claude Code sessions through OpenClaw:

# Continue last session
claude -c

# Resume specific session
claude -r my-session "Continue working on the feature"

Method 3: MCP Integration

For advanced integration, configure MCP servers:

claude mcp add <server-name> <config>

Examples

Example 1: Explore a New Codebase

/claude-code Explore the project structure and tell me what the main components are

Example 2: Fix a Bug

/claude-code Find and fix the login bug - users can't log in with correct credentials

Example 3: Write Tests

/claude-code Write unit tests for the payment module

Example 4: Refactor Code

/claude-code Refactor the auth middleware to use async/await

Example 5: Code Review

/claude-code Review this PR for security issues

Example 6: Batch Operations

/claude-code /batch migrate all API endpoints to use the new error handler

Troubleshooting

Authentication Issues

# Check auth status
claude auth status

# Re-authenticate
claude auth login

Permission Issues

# Run with specific permission mode
claude --permission-mode plan "task"

Session Issues

# List available sessions
claude --resume

# Fork a session (create new)
claude -r old-session --fork-session "new task"

Advanced Features

Sub-agents

Run multiple Claude Code instances in parallel:

claude --agents '{
  "reviewer": {
    "description": "Code reviewer",
    "prompt": "You are a senior code reviewer",
    "tools": ["Read", "Grep", "Bash"]
  }
}'

Hooks

Automate actions before/after Claude Code runs:

See Hooks documentation

MCP Servers

Connect Claude Code to external tools:

claude mcp add github '{"github_token": "..."}'

Resources

• Official Docs
• CLI Reference
• Skills Guide
• MCP Documentation
• Common Workflows

---

Note: Claude Code requires authentication. Run claude auth login before first use.