How to Invoke the clone-website Skill with Target URLs in JCodesMore/ai-website-cloner-template

You invoke the /clone-website skill by typing the command followed by one or more URLs directly into your AI coding agent's chat interface after installing the project dependencies.

The JCodesMore/ai-website-cloner-template repository provides a built-in AI skill that automates website cloning through a multi-phase pipeline. When you invoke the clone-website skill with target URLs, the tool validates each address, creates dedicated output folders, and executes a complete cloning workflow from reconnaissance to component generation.

Prerequisites: Project Setup and Agent Launch

Before invoking the skill, you must initialize the project and start a compatible AI coding agent.

First, install the dependencies:

npm install

Then launch an AI agent that reads the AGENTS.md instructions. Supported agents include Claude Code, Codex CLI, OpenCode, and Cursor. For example, with Claude Code:

claude --chrome

The agent automatically loads the skill definitions from .github/skills/clone-website/SKILL.md and the central instructions in AGENTS.md, making the /clone-website command available in the chat session.

Invoking the Skill with Target URLs

The canonical syntax accepts one or more space-separated URLs as documented in the README.md at line 49:

/clone-website <target-url1> [<target-url2> ...]

The skill definition in .github/skills/clone-website/SKILL.md (line 4) confirms this argument pattern with the hint: argument-hint: "<url1> [<url2> ...]".

Single URL Execution

To clone a single website, enter the command followed by the full URL:

/clone-website https://www.vercel.com

The skill validates the URL, creates a folder at docs/research/<hostname>/, and launches the full cloning pipeline.

Multiple URLs Parallel Processing

For batch operations, pass multiple URLs separated by spaces. The skill processes these in parallel:

/clone-website https://nextjs.org https://tailwindcss.com https://shadcn.dev

Each URL receives its own isolated output directory under docs/research/, preventing conflicts between concurrent cloning operations.

The Cloning Pipeline Execution

After you invoke the command with valid target URLs, the skill executes a multi-phase pipeline:

1. URL Validation – Each argument is checked for valid URI formatting and accessibility.
2. Directory Creation – The system generates per-site folders at docs/research/<hostname>/ to store research data and generated components.
3. Pipeline Execution – The skill runs through reconnaissance → foundation → component specs → parallel builds.

The scripts/sync-skills.mjs file regenerates platform-specific copies of the skill definition, ensuring consistent behavior across Claude, Codex, Cursor, and other supported agents.

Summary

• Invoke the skill by typing /clone-website followed by one or more URLs in your AI agent chat.
• Reference the canonical syntax in README.md (line 49) and the skill definition in .github/skills/clone-website/SKILL.md (line 4).
• Process multiple URLs in parallel by separating them with spaces; each gets its own docs/research/<hostname>/ output folder.
• Require prior installation (npm install) and a running AI agent that supports the AGENTS.md protocol.

Frequently Asked Questions

How many URLs can I pass to the clone-website skill at once?

The skill accepts multiple URLs as space-separated arguments, and the underlying pipeline processes them in parallel. While there is no explicit hard limit in the source code, each URL creates its own docs/research/<hostname>/ directory, so execution time scales with the number of targets and site complexity.

What happens if I provide an invalid URL?

The skill validates each URL before initiating the cloning pipeline. Invalid or malformed URLs fail early in the validation step, preventing the creation of empty output directories and ensuring the pipeline only runs against accessible targets.

Where are the cloned website files stored?

Output files are organized under docs/research/<hostname>/ where <hostname> is derived from each target URL. This structure keeps research data, component specs, and generated code isolated per site, allowing you to clone multiple properties without file conflicts.

Do I need to restart the AI agent between cloning sessions?

No. Once the AI agent loads the AGENTS.md instructions and skill definitions from .github/skills/clone-website/SKILL.md, the /clone-website command remains available for the duration of the session. You can invoke the skill multiple times with different URLs without restarting the agent.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}