AI Tools Setup Assistant
Install, configure, and repair Claude Desktop, MCP servers, and local AI tooling on your own machine.
Evidence shows the skill explicitly instructs backing up before editing, never echoing secrets, verifying before stopping, and advising users to trust MCP sources, demonstrating security awareness and least-privilege practice. However, the underlying Desktop Commander MCP is designed to execute arbitrary terminal commands, and its security model explicitly states its restrictions are not a sandbox, placing responsibility on the user. Deductions: incomplete permission control, unclear user confirmation process, and high-risk dependency capabilities.
The skill provides detailed diagnostic steps and concrete commands, showing good organization, but lacks self-contained tests for the skill itself and relies on external documentation without verification. Static assessment cannot execute commands or confirm key paths, limiting the score. Deductions: no test coverage for key paths, insufficient error handling feedback, execution unverified.
The skill covers multiple AI tooling scenarios (Claude Desktop, MCP, OpenClaw, Hermes) with clear trigger conditions, but boundary and non-fit descriptions are limited. Environment fit is not explicitly assessed, and the skill appears English-focused with no Chinese support addressed. Deductions: incomplete boundaries, no Chinese support consideration.
The skill provides detailed reference files and installation instructions, follows a clear information hierarchy, and includes known limitations (e.g., security model in SECURITY.md). However, version is only 0.1.0 and no changelog is present. Deductions: incomplete versioning, missing changelog.
The skill offers detailed step-by-step guidance to accomplish configuration and repair tasks, but output format and direct usability are unverified. Static assessment cannot confirm actual outcomes. Deductions: output completeness unverified, lack of direct usability evidence.
The skill cites external documentation and official resources, but lacks independent execution evidence. Repository contains test files, but they focus on core MCP functionality, not this skill specifically. Deductions: no independent reproducible evidence for key claims of the skill.
- The underlying Desktop Commander MCP can execute arbitrary commands and is not a sandbox; users must ensure their AI client is trusted and uncompromised.
- External documentation referenced (e.g., OpenClaw, Hermes) may be unreachable from mainland China networks, affecting practical use.
- Static assessment did not verify actual operation; test in a controlled environment before production use.
What does this skill do, and when should you use it?
This skill guides an AI agent to help users set up, connect, validate, and repair Claude Desktop and MCP servers using Desktop Commander's file and terminal tools. It focuses on configure-and-troubleshoot tasks rather than clean installs, covering config files like mcp.json and claude_desktop_config.json. The skill also covers local agent gateways such as OpenClaw and Hermes, providing a systematic workflow from detection to verification. It's aimed at users facing common issues like 'Claude can't see my tools' or 'my MCP server isn't showing up'.
This skill performs the following operations:
- Detects the OS and shell by reading config and listing processes.
- Verifies model/API keys with a quick test call using start_process.
- Reads config files like mcp.json / claude_desktop_config.json using read_file.
- Edits config files with edit_block, backing up first with start_process (cp to .bak).
- Validates JSON after edits by parsing with start_process.
- Verifies processes and ports with list_processes and start_process.
- Routes to reference files: claude-desktop-mcp.md, openclaw.md, or hermes.md based on the task.
- A user reports 'MCP server isn't showing up' or 'tools aren't loading in Claude.'
- A user wants to add a new MCP server to Claude Desktop.
- A user needs to fix a broken claude_desktop_config.json file.
- A user wants to wire up local model tools like Ollama or LM Studio.
- A user needs to diagnose a local AI agent gateway like OpenClaw or Hermes.
- A user is facing 'agent won't respond' errors and needs step-by-step troubleshooting.
What are this skill's strengths and limitations?
- Specialized in troubleshooting common MCP connection issues.
- Provides a clear step-by-step workflow and golden rules.
- Covers multiple tools (Claude Desktop, OpenClaw, Hermes).
- Emphasizes backups and verification to avoid data loss.
- Stresses security (no secret echo, redaction).
- Depends on Desktop Commander MCP server; won't work without it.
- Only provides guidance via reference files; no actual automation scripts.
- Platform-specific paths and commands (macOS/Windows/Linux) may be incomplete.
- Reference files may not be consistently maintained.
- No test suite or official validation mentioned.
How do you install this skill?
This skill is part of the Desktop Commander MCP repository. Install the Desktop Commander MCP server (e.g., via npx @wonderwhy-er/desktop-commander@latest setup or manually add to claude_desktop_config.json), then use the skill within Claude Desktop or a compatible MCP client. The skill itself requires no separate installation; it leverages the file/terminal tools from the connected MCP server.
How do you use this skill?
In Claude Desktop, when facing issues related to MCP configuration or local AI tools, describe the symptom, e.g., 'My MCP server isn't connecting' or 'Fix my Claude Desktop config.' The skill will guide the agent through detection, configuration, verification, and repair steps. For example, say 'Set up Ollama' or 'Diagnose my OpenClaw gateway.'