openclaw-claude-code_skill

This skill lets you control Claude Code via MCP to execute commands, read and write files, and manage agent teams efficiently.

Python

2.6k

GitHub Stars

6

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill openclaw/skills --skill openclaw-claude-code

_meta.json291 B
mcp_config.example.json648 B
package.json679 B
README.md6.9 KB
SKILL.md12.0 KB
tsconfig.json399 B

Overview

This skill lets you control Claude Code via the MCP (Model Context Protocol) from scripts, CLIs, or OpenClaw agents. It supports persistent sessions, agent teams, direct tool calls, streaming output, permission modes, and budget limits. Use it to automate multi-step development workflows, code review, testing, and complex agent collaboration.

How this skill works

The skill connects to a Claude Code MCP endpoint (or a proxy) and exposes all Claude Code tools programmatically: Bash, Read, Write, Edit, Glob, Grep, Task, WebFetch, Git, and more. You can run single direct tool calls or start persistent sessions that maintain context, allow agent teams, stream progress, preview edits, and enforce tool whitelists/blacklists and budgets. Sessions can be paused, forked, resumed, inspected, and searched.

When to use it

Multi-step workflows that need state and iterative edits (use persistent sessions)
Autonomous agents that must call multiple tools and maintain context
Quick one-off operations like reading or running a single command (use direct tools)
Code reviews, automated testing, or refactors where streaming feedback is useful
Coordinated work across specialized agent roles (architect, developer, reviewer)

Best practices

Prefer persistent sessions for multi-step tasks to avoid reconnect/reload cycles
Use --permission-mode plan to preview changes before applying them
Set --max-budget to limit API spending and prevent runaway sessions
Whitelist only the tools needed for the task; disallow dangerous commands (rm, sudo)
Fork sessions before risky experiments and use resume/fork to branch work

Example use cases

Start a session to iteratively refactor a module, run tests, and apply fixes with streaming logs
Run a security-focused agent team to scan TypeScript files and produce prioritized findings
Execute single tool calls to read files, run npm tests, or search for TODOs across a repo
Automate test generation: find untested functions, write tests, run suite, and fix failures
Use a proxy to route requests to other models (Gemini/GPT) while maintaining the Claude Code workflow

FAQ

Yes. Use --allowed-tools and --disallowed-tools to whitelist and blacklist tools or patterns.

How do I avoid unexpected edits?

Start sessions with --permission-mode plan to preview edits, or use acceptEdits only when safe.