open-supermarkets: Instructions file for Codex

Agent Integration Guide

This document explains how to integrate Sainsbury's CLI into AI agent frameworks.

---

Supported Frameworks

• ✅ OpenClaw / Clawdbot - Skills system
• ✅ Pi Agent / Mom - Slack bot with skills
• ✅ Claude Desktop - MCP server (future)
• ✅ Custom agents - Any framework that can call bash

---

Quick Integration

1. Add as Skill

Copy to your agent's skills directory:

cp -r sainsburys-cli /path/to/agent/skills/

2. Agent Calls Commands

// From your agent code
await bash("cd skills/sainsburys-cli && npm run groc search 'milk'");

3. Parse JSON Responses

const stdout = await bash("cd skills/sainsburys-cli && npm run groc search 'milk' --json");
const results = JSON.parse(stdout);

results.products.forEach(product => {
  console.log(`${product.name} - £${product.retail_price.price}`);
});

---

Skill File Format

The SKILL.md follows the open skills format used by OpenClaw, Pi, and other agent frameworks.

Frontmatter

---
name: sainsburys-groceries
description: AI-powered meal planning and grocery ordering
license: MIT
compatibility: Node.js 18+, TypeScript, Playwright
metadata:
  author: zish
  version: "2.0.0"
allowed-tools: Bash({baseDir}/node:*), Bash(npm:run:groc:*)
---

Triggers

Agent should load this skill when user:

• Mentions meal planning or groceries
• Asks "what's for dinner?"
• Wants to order food/shopping
• Talks about recipes or cooking
• Mentions Sainsbury's

---

Natural Language Workflow

User Intent Detection

const intents = {
  mealPlanning: ["plan meals", "what should I cook", "dinner ideas"],
  shopping: ["add to basket", "order groceries", "buy milk"],
  delivery: ["book slot", "delivery Tuesday", "checkout"],
  query: ["what's in my basket", "show orders", "search for bread"]
};

if (userMessage.match(/plan meals|what.*cook|dinner ideas/i)) {
  await startMealPlanning();
}

Meal Planning Flow