Integration Architect Skill Suite 🚀

An autonomous, multi-skill suite for the Integration Architect lifecycle. It bridges the gap between legacy technical systems and modern mobile frontend requirements using a Compute-First, Deterministic approach.

🛠 Skills Included

1. integration-architect: The Master Orchestrator. Manages the 5-phase lifecycle.
2. discovery-expert: Technical Source understanding (Word/PowerPoint/Excel/CSV to OpenAPI/Schema Map).
3. ui-analyzer: User Target understanding (Figma/Screenshot to UI Spec).
4. contract-designer: The Bridge (Optimized OpenAPI & TypeScript Types).
5. data-aligner: Persistence Layer (DB Gap Analysis & SQL Alignment).
6. quality-architect: Verification Layer (Postman, K6, and Mocks).

📦 Installation

Option A: Claude Code Plugin (Recommended)

Install directly within Claude Code using the /plugin command:

/plugin install https://github.com/pongsatorna/integration-architect-agent

After installation, use the /integrate slash command to start the lifecycle:

/integrate              # Run full 5-phase lifecycle
/integrate discovery    # Start from a specific phase

Figma MCP Setup: Set your Figma token as an environment variable:

export FIGMA_ACCESS_TOKEN="your-figma-personal-access-token"

Option B: Antigravity CLI (Gemini)

To install the entire suite of skills as an Antigravity plugin:

agy plugin install https://github.com/pongsatorna/integration-architect-agent

Option C: Individual Skills (Manual)

Clone the repository and symlink specific skills to your local workspace:

git clone https://github.com/pongsatorna/integration-architect-agent.git
mkdir -p .agents/skills
ln -s $(pwd)/integration-architect-agent/skills/integration-architect .agents/skills/
ln -s $(pwd)/integration-architect-agent/skills/discovery-expert .agents/skills/
# etc...

⚙️ Prerequisites

These skills are Self-Healing. The agent will automatically attempt to install the following if missing:

• Python 3.12+
• Libraries: pandas, mammoth, python-docx, python-pptx, openpyxl, pyyaml

Figma MCP Setup

To enable high-fidelity UI analysis, you need to configure the Figma MCP Server.

For Claude Code (Option A): The .mcp.json is bundled with the plugin — just set the environment variable:

export FIGMA_ACCESS_TOKEN="your-figma-personal-access-token"

For Antigravity/Gemini (Option B): Since the Antigravity CLI (agy) scans the workspace root, this configuration must be placed in .agents/mcp_config.json at the root of your project (not inside ./discovery/).

Create the file .agents/mcp_config.json in the root of your project:

{
  "mcpServers": {
    "figma": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-figma"
      ],
      "env": {
        "FIGMA_ACCESS_TOKEN": "YOUR_FIGMA_PERSONAL_ACCESS_TOKEN"
      }
    }
  }
}

Note: Replace YOUR_FIGMA_PERSONAL_ACCESS_TOKEN with your actual Figma Personal Access Token.

📂 Workspace Structure

User-Created Input Folders (Must be set up before running):

• ./discovery/: Place your legacy API docs (.docx, .pptx, .xlsx) and database schema exports (.csv) here.
• ./ui_analysis/screenshots/: Place UI mockups/screenshots here (if not using the Figma MCP server).

Agent-Created Output Folders (Created automatically at runtime):

• ./contracts/: Final OpenAPI specs, TypeScript types, and lineage maps.
• ./database/: SQL alignment scripts and database migration plans.
• ./testing/: Postman collections, K6 load test scripts, and mock response JSONs.

---

Build with ❤️ for Integration Architects.