← Back to Skills Marketplace
5433
Downloads
5
Stars
32
Active Installs
1
Versions
Install in OpenClaw
/install openclaw-claude-code
Description
Programmatically control Claude Code via MCP protocol to execute commands, manage sessions, search and edit code, and coordinate agent teams for development...
Usage Guidance
Install only if you intend to give this skill Claude Code-level control over selected projects. Use plan or default permission modes for sensitive work, avoid bypassPermissions/skip-permissions, restrict allowed tools and added directories, run it with a minimal environment that does not contain unrelated secrets, use only trusted and pinned MCP servers, keep backend URLs local or controlled by you, and stop persistent sessions when finished.
Capability Analysis
Type: OpenClaw Skill
Name: openclaw-claude-code
Version: 0.3.0
The skill is classified as suspicious due to its extensive high-risk capabilities, which, while potentially necessary for its stated purpose, could be leveraged for malicious activities if misused or compromised. Key indicators include the ability to execute arbitrary shell commands (`claude-code-skill bash <command>`) and launch arbitrary processes via MCP servers (`src/mcp/client.ts`, `mcp_config.example.json`), arbitrary file system read/write access (`read`, `Write` tools in `src/index.ts`), and the option to bypass security permissions (`--skip-permissions` in `src/index.ts`, `bypassPermissions` mode in `SKILL.md`). Furthermore, the `--base-url` option allows the agent to connect to arbitrary external API endpoints, enabling potential data exfiltration or unauthorized actions if directed to a malicious server, as documented in `SKILL.md` and `README.md`.
Capability Assessment
Purpose & Capability
The core purpose is coherent and disclosed: control Claude Code via MCP, including Bash, read/write tools, raw tool calls, persistent sessions, and agent teams. The concern is the breadth of authority: direct shell execution, arbitrary Claude Code tool invocation, file mutation, custom model/backend routing, and agent delegation are all available from the skill surface.
Instruction Scope
The documentation mentions plan mode and labels bypassPermissions/skip-permissions as dangerous, but high-impact modes and flags are still normal CLI options. The default session permission mode is acceptEdits, raw bash/call commands have no local confirmation barrier, and system-prompt/tool/access-directory options can materially change behavior.
Install Mechanism
The install path is disclosed as a local Node-based package, with README steps to clone, npm install, build, and optionally link globally. Example MCP configurations use npx to fetch MCP servers and include token-bearing integrations, so users should pin and trust those server packages separately.
Credentials
The default backend is local, but the CLI sends prompts, commands, and session actions over HTTP to a configurable backend URL. More importantly, the MCP client forwards nearly the entire parent process environment to spawned MCP server processes, which can expose unrelated API keys or credentials to any configured server.
Persistence & Privilege
Persistent sessions, history, search, resume, fork, pause, and multi-agent workflows are purpose-aligned and disclosed, but they retain context and can continue operating with broad tool permissions until stopped or constrained by the user.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openclaw-claude-code - After installation, invoke the skill by name or use
/openclaw-claude-code - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.3.0
Major update with comprehensive documentation and advanced features.
- Added full SKILL.md guide covering installation, usage, command reference, and advanced options.
- Detailed instructions for managing persistent sessions, direct tool calls, and agent teams.
- Descriptions of permission modes, tool whitelisting, and contextual session management.
- Included examples for batch operations, session search/history, and agent collaboration.
- Clarified multi-model support and custom configuration for expert users.
Metadata
Frequently Asked Questions
What is Claude Code Agent?
Programmatically control Claude Code via MCP protocol to execute commands, manage sessions, search and edit code, and coordinate agent teams for development... It is an AI Agent Skill for Claude Code / OpenClaw, with 5433 downloads so far.
How do I install Claude Code Agent?
Run "/install openclaw-claude-code" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Claude Code Agent free?
Yes, Claude Code Agent is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Claude Code Agent support?
Claude Code Agent is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Claude Code Agent?
It is built and maintained by Enderfga (@enderfga); the current version is v0.3.0.
More Skills