Does the clone-website Skill Have an API? Understanding AI-Agent Commands vs REST Endpoints

The clone-website skill does not expose a REST or GraphQL API; it is a command-style AI-agent skill invoked via slash commands (e.g., /clone-website <url>) within supported coding platforms.

The JCodesMore/ai-website-cloner-template repository implements website cloning as an internal pipeline orchestrated by AI-coding agents, not as a traditional web service. All processing occurs within the Next.js project and the agent runtime, meaning you cannot trigger it with a fetch request or curl command from an external client.

What Is the clone-website Skill?

The clone-website capability is a declarative skill definition designed for AI-coding agents such as GitHub Copilot, Claude Code, or Gemini CLI. Unlike conventional APIs that expose HTTP endpoints, this skill operates as a chat-integrated command that parses user intent from natural language or slash syntax and executes a multi-phase internal workflow.

When you type /clone-website https://example.com in a compatible interface, the agent runtime loads the command definition and triggers a sequence of local scripts that handle reconnaissance, asset extraction, component generation, and final assembly.

Why There Is No External REST API

The architecture deliberately avoids external network endpoints for security and integration reasons. According to the repository structure, the skill executes entirely within the host project's environment:

• No exposed ports or route handlers accept external HTTP requests for cloning
• Processing happens inside the Next.js application context and AI-agent runtime
• State management relies on local file system operations (writing to docs/research/, src/app/page.tsx, etc.) rather than API response payloads

If you attempt to locate a pages/api/clone.ts or similar endpoint, you will not find one. The skill's entry points are the command definitions parsed by the AI agent itself.

How the Skill Is Defined and Distributed

The repository uses a sync-and-propagate pattern to maintain skill definitions across multiple AI platforms.

The Master Skill File

The canonical definition lives at .github/skills/clone-website/SKILL.md. This YAML-style markdown file contains:

• Human-readable description and argument schemas
• Phase-by-phase workflow definitions (reconnaissance, foundation build, component dispatch)
• Tool permissions (browser MCP, file system access)

Platform-Specific Command Generation

The scripts/sync-skills.mjs Node.js script reads the master skill and auto-generates platform-specific command files:

// Regenerate all platform command files after editing SKILL.md
node scripts/sync-skills.mjs

This creates:

• .opencode/commands/clone-website.md for OpenAI/Claude-style agents
• .gemini/commands/clone-website.toml for Gemini agents
• Additional formats in .github/skills/ subdirectories

These generated files tell each AI runtime how to parse the slash command and which internal scripts to execute.

How to Invoke the clone-website Skill

Since no HTTP API exists, interaction follows two distinct patterns depending on your environment.

Using Slash Commands in AI Coding Platforms

In supported environments (VS Code with AI extensions, Claude Code terminal, etc.), simply type:

/clone-website https://example.com https://another-site.com

The agent runtime:

1. Parses the command against the auto-generated definition
2. Validates URLs against the skill's argument schema
3. Executes the internal pipeline phases sequentially

Running the Pipeline Locally

To trigger the same workflow programmatically without an AI chat interface, you must manually invoke the underlying Node.js scripts referenced in the skill definition:

# Install dependencies first

npm install

# Execute asset discovery (Phase 2 of the skill)

node scripts/download-assets.mjs https://example.com

Note: This requires reproducing the full phase sequence manually—reconnaissance (Phase 1), foundation build (Phase 2), component spec generation (Phase 3), parallel builder dispatch (Phase 4), and assembly (Phase 5)—as defined in .github/skills/clone-website/SKILL.md.

The Internal Pipeline Architecture

When invoked, the skill orchestrates five distinct phases using local scripts and MCP (Model Context Protocol) tools:

1. Reconnaissance – Browser automation captures screenshots and extracts design tokens
2. Foundation Build – Updates global styles and downloads assets to public/
3. Component Spec Generation – Writes exhaustive markdown specs to docs/research/components/
4. Parallel Builder Dispatch – Spins up git worktrees for concurrent component construction
5. Assembly & QA – Merges components into src/app/page.tsx and performs visual diffing

All file paths, including src/app/page.tsx and scripts/download-assets.mjs, are referenced relative to the project root, confirming the skill operates as a local build process rather than a remote service.

Summary

• The clone-website skill is not a REST API; it is a command-layer integration for AI coding agents.
• Invocation happens via slash commands parsed by the agent runtime, not HTTP requests.
• Definitions are stored in .github/skills/clone-website/SKILL.md and propagated by scripts/sync-skills.mjs.
• Local execution requires manually running the Node.js scripts found in the scripts/ directory.
• All processing is internal to the Next.js project, utilizing the local file system and MCP tools.

Frequently Asked Questions

Can I call the clone-website skill from my own backend?

No. There is no exposed endpoint, webhook, or serverless function that accepts external requests. To trigger cloning from a backend service, you would need to execute the Node.js scripts in scripts/ (such as download-assets.mjs) directly within a Node environment that has access to the repository files.

What file formats define the clone-website command?

The primary definition is a YAML-style markdown file at .github/skills/clone-website/SKILL.md. The scripts/sync-skills.mjs script transpiles this into various formats including Markdown (.opencode/commands/clone-website.md) and TOML (.gemini/commands/clone-website.toml) to satisfy different AI agent runtime requirements.

How do I update or customize the skill behavior?

Edit the master skill file at .github/skills/clone-website/SKILL.md, then run node scripts/sync-skills.mjs to regenerate platform-specific command definitions. Changes to workflow phases should be followed by testing the individual scripts in the scripts/ directory to ensure compatibility with your modifications.

Is there a way to trigger cloning without an AI coding agent?

Yes, but it requires manual execution. You can run the underlying scripts like scripts/download-assets.mjs directly from your terminal. However, you must replicate the full pipeline sequence—reconnaissance, asset processing, spec generation, and component assembly—as the AI agent normally handles this orchestration automatically when you use the slash command.

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