{
  "id": 15815124,
  "title": "Our coding agent prompts live in the ticket, reviewed like PRs",
  "url": "https://urgent.news/2026/10/20/our-coding-agent-prompts-live-in-the-ticket-reviewed-like-prs",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-20T23:00:29.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/panthpatel/our-coding-agent-prompts-live-in-the-ticket-reviewed-like-prs-4f2g"
  },
  "original_language": "en",
  "account": "Panth leads the software team at Oizom, an organization that uses Envizom to turn data from air-quality and environmental monitors into usable information for cities and industries. The key design goal for their coding agent pipeline is to keep the prompt stored with the work, making it possible to review and coach the way people ask for things. When a developer writes a prompt in plain English and drags it to Approved, the ticket becomes the only channel for communication with the agent. The orchestrator, orch, creates a separate stack for each ticket, including a branch in every repository, backend, drivers, simulator, frontend, a copy of the development database, and a private preview URL. The coding agent performs the work within this stack, and when finished, posts the preview link back into the ticket thread, moving it to QAQC. Developers can then test the preview, provide feedback through the same thread, and move the ticket back to the next stage. The entire exchange, from the initial prompt to the agent's output and the developer's feedback, is stored within the thread, creating a self-contained record of the entire process. The prompts, results, and review trail all coexist within the same ticket, providing an audit trail that shows what was asked, built, tested, and the feedback received. By storing prompts with the work they produce, Oizom's team can review and coach the way their members ask for things, gaining insights that go beyond generic prompting guides. The cost of each agent turn is also visible within the thread, influencing how future prompts are written. Once the ticket is moved to Git Push, the agent commits, merges, opens, and merges the pull requests, waits for CI, deploys to the development environment, checks its health, and posts the PR and commit links back into the thread. This final step completes the entire pipeline, with all relevant information contained within the ticket.",
  "summary": "Every prompt my team writes to our coding agent is stored in the ticket that asked for it. The ticket thread is the only way a developer can talk to the agent, so there is nowhere else for a prompt to go. I'm Panth, and I lead the software team at Oizom, where Envizom is the platform that turns data from air-quality and environmental monitors into something cities and industries can act on. This…",
  "key_points": [
    "Panth leads software team at Oizom using Envizom for data processing.",
    "Coding agent pipeline stores prompts with work for review and coaching.",
    "Ticket serves as communication channel and audit trail for entire pipeline."
  ],
  "editors_take": null,
  "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."
}