{
  "id": 16401412,
  "title": "Trojan Horse: AI agents that call mechanics for you, and how Kane CLI caught our bug",
  "url": "https://urgent.news/2026/10/25/trojan-horse-ai-agents-that-call-mechanics-for-you-and-how-kane-cli",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-25T00:48:25.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/chandrakhushi/trojan-horse-ai-agents-that-call-mechanics-for-you-and-how-kane-cli-caught-our-bug-10p0"
  },
  "original_language": "en",
  "account": "In the dark hours after a car malfunction, drivers often scramble to find a mobile mechanic. Our team tackled this challenge at the recent Kane CLI Hack Day in San Francisco, crafting a multi-agent app they dubbed \"Trojan Horse\" (\"Engine cooked? Help booked.\"). The app, built in a single day, consists of a Discovery Agent that uses OpenAI web search to identify nearby mechanics, a Caller Agent that dials up to three mechanics concurrently via OpenAI Realtime and Vonage/Twilio, and a Callback Agent that confirms the booking once the mechanic confirms availability, arrival time, and price. The finished product is a React dashboard with a Node orchestrator and SQLite database, featuring separate discovery and voice services backed by shared Zod contracts.\n\nTo ensure the app's integrity, the team put it through its paces with Kane CLI, which, given the app's specifications, generated 26 acceptance criteria and 19 requirement-linked browser tests. The tests included scenarios where mechanics intentionally behaved incorrectly—such as failing to respond, providing unknown prices or estimated arrival times, arriving late, or failing to confirm the booking. These tests were designed to be repeatable, and Kane CLI automatically replays the test suite on every pull request.\n\nIn one critical test, Kane CLI simulated a mechanic who hesitated, saying, \"I'm not sure anyone is free right now.\" The app, designed to display \"Unknown\" if the mechanic did not confirm availability, mistakenly displayed \"Available now: No.\" This error would have misled a driver, potentially leading them to the wrong conclusion about the mechanic's availability. In response, the team adjusted their shared contract to make the availability field nullable, thereby allowing for \"Unknown\" states. The updated test now passes, ensuring that the app accurately reflects the mechanic's true availability. The evidence from these tests is available in the public repository, demonstrating the app's robustness and reliability. For more information and access to the codebase, visit https://github.com/RamyaRamachandran-driod/emergency-caller.",
  "summary": "The problem Your car breaks down at night. You search for mobile mechanics and call them one by one: who’s open, who’s close, who won’t overcharge? That’s exactly when time matters most. What we built Trojan Horse (“Engine cooked? Help booked.”) is a multi-agent app we built in one day at Kane CLI Hack Day SF: A Discovery Agent uses OpenAI web search to shortlist mobile mechanics near you. A…",
  "key_points": [
    "Trojan Horse AI app connects drivers with mechanics via mobile",
    "Discovery, Caller, and Callback agents handle booking process",
    "Kane CLI tests reveal bug where app incorrectly showed mechanic as available"
  ],
  "editors_take": "Integrating automated acceptance criteria into the development cycle allows multi-agent applications to handle the unpredictability of human interactions without providing misleading information to end users.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}