How to Add Support for New AI Coding Agent Platforms in ai-website-cloner-template

To add support for a new AI coding agent platform in ai-website-cloner-template, extend either scripts/sync-agent-rules.sh for rule-based agents or scripts/sync-skills.mjs for skill-based agents, then regenerate configuration files from the single source of truth in AGENTS.md .

The ai-website-cloner-template repository maintains all AI agent instructions in a central AGENTS.md file. Through automated synchronization scripts, the template propagates updates to platform-specific configurations without duplicating content, ensuring that changes to the core prompt automatically reach every supported coding agent.

Understanding the Architecture

The template categorizes AI coding agents into two distinct integration patterns based on how they consume instructions:

Rule-Based Platforms

Agents like Continue and Amazon Q expect plain-text rule files—typically markdown documents—that are fed verbatim to the model as system instructions. These platforms read static files containing the complete prompt without dynamic argument substitution.

Skill-Based Platforms

Agents like Codex CLI, GitHub Copilot, and Gemini CLI consume structured skill definitions that may include placeholders for arguments. These definitions typically use JSON, TOML, or specialized markdown formats that support dynamic variable injection.

Step-by-Step Implementation

1. Determine the Platform Type

Identify whether your target agent requires a rule-based file (plain markdown/text read as-is) or a skill-based definition (structured format with argument placeholders). This decision determines which synchronization script you must modify.

2. Extend the Appropriate Sync Script

For rule-based platforms, append a new write_file call to scripts/sync-agent-rules.sh:

# My-Agent — .myagent/rules/project.md

write_file "$REPO_ROOT/.myagent/rules/project.md" "$RESOLVED_CONTENT"

For skill-based platforms, add a new write call to scripts/sync-skills.mjs:

// My-Agent — JSON skill definition
write(
  '.myagent/cli-agents/clone-website.json',
  JSON.stringify(
    {
      name: 'clone-website',
      description: shortDesc,
      prompt: noArgs(body),
      fileContext: ['AGENTS.md'],
    },
    null,
    2
  ) + '\n'
)

The write_file helper automatically prepends an auto-generated header and handles directory creation. In the JavaScript script, write creates missing directories and logs confirmation, while noArgs transforms $ARGUMENTS placeholders for platforms that do not support dynamic inputs.

3. Update Documentation

Add the new platform to the Supported Platforms table in README.md. If the platform requires specific command invocations, document these in the Commands section to assist users in discovering the integration.

4. Regenerate Configuration Files

Execute the synchronization scripts to generate the new platform's files:

bash scripts/sync-agent-rules.sh

node scripts/sync-skills.mjs

These scripts write generated files to appropriate hidden directories (e.g., .myagent/rules/ or .myagent/cli-agents/) while preserving the DRY principle—AGENTS.md remains the sole editable source.

5. Verify Generated Output

Open the generated files to confirm:

• The header <!-- AUTO-GENERATED ... --> is present in rule-based files.
• Placeholder substitutions (such as $ARGUMENTS or {{args}}) match the target platform's expected syntax.
• JSON or TOML skill files validate against the agent's schema requirements.

6. Commit Your Changes

git add scripts/sync-agent-rules.sh scripts/sync-skills.mjs .myagent/
git commit -m "Add My-Agent support for clone-website command"

Key Implementation Details

The scripts/sync-agent-rules.sh script resolves content from AGENTS.md and uses the write_file function to output plain-text rules with automatic header injection. Conversely, scripts/sync-skills.mjs parses the same source document but structures the output as named skills using the write function, which supports multiple output formats including Markdown, TOML, and JSON. Both scripts ensure that updates to AGENTS.md propagate identically across all supported platforms.

Summary

• The repository uses AGENTS.md as the single source of truth for agent instructions.
• Rule-based integrations require extending scripts/sync-agent-rules.sh to write plain-text files.
• Skill-based integrations require extending scripts/sync-skills.mjs to generate structured definitions.
• Helper functions handle directory creation, header injection, and argument substitution automatically.
• Always verify generated files for correct headers and placeholder syntax before committing.

Frequently Asked Questions

What is the difference between rule-based and skill-based agents?

Rule-based agents consume plain markdown or text files containing system instructions verbatim, while skill-based agents use structured definitions (JSON, TOML, or specialized Markdown) that define named capabilities with support for dynamic argument placeholders.

Do I need to modify AGENTS.md when adding a new platform?

No. AGENTS.md serves as the immutable source of truth. You only need to extend the synchronization scripts to format and export its content to new platform-specific file locations and formats.

How do I handle argument placeholders for different platforms?

Use the noArgs function in scripts/sync-skills.mjs to strip or transform $ARGUMENTS placeholders, or implement custom substitution logic that converts template variables into the specific syntax required by your target platform's skill definition schema.

Where should generated files be placed for new agents?

Follow the existing convention: place rule-based files under .myagent/rules/ and skill-based files under .myagent/cli-agents/, mirroring the structure used for existing platforms like .github/copilot/skills/ or .continue/rules/.

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