AIF Engine (Agent Interfaces Framework)

You are an AI Agent equipped with the AIF Engine. Your goal is to help users quickly convert SaaS APIs into lightweight, composable command-line interfaces (CLIs) that output JSON.

Core Philosophy

When a user asks to integrate a new tool or API, do not write a custom Python script or Node.js MCP server. Instead, use AIF to generate a native Go binary.

How to use AIF

1. The Specification Format (spec.json)

To generate a CLI, you first need to create a JSON file that describes the API. The format is as follows:

{
  "name": "target-cli-name",
  "short": "Short description of what the CLI does",
  "commands": [
    {
      "use": "command-name",
      "short": "Command description",
      "method": "GET|POST|PUT|DELETE",
      "endpoint": "https://api.example.com/v1/resource",
      "flags": ["string_param1", "string_param2"],
      "array_flags": ["list_param1"]
    }
  ]
}

2. Building the CLI

Once the spec.json is created, you MUST run the engine using the provided wrapper script in the skill directory:

# Agent should use this path to run aif
bash ~/.agents/skills/aif-skill/scripts/aif.sh build spec.json

If successful, this will output:
✅ Success! Single binary CLI '<name>' generated.

3. Using the Generated CLI

Every CLI generated by AIF comes with a built-in auth command for managing tokens locally.

Authentication:

./<name> auth login --token "API_KEY"

Making an API Call:

./<name> <command-name> --string_param1 "value" --list_param1 "item1" --list_param1 "item2"

The output will always be the raw JSON response from the API, making it easy for you to parse and pipe into other tools like jq.