What is chatgpt-task-hook?

This is a local macOS tool that lets you trigger ChatGPT or Codex tasks through an HTTP API instead of scheduling them at fixed times. The author built it for personal use because the normal task scheduling in ChatGPT only supports time-based triggers, not event-based ones. The project provides a web interface styled to match ChatGPT's existing scheduled tasks panel, and it runs alongside the ChatGPT desktop app. Each task stores a prompt that gets sent to Codex or ChatGPT, with configurable options for the model, reasoning effort, project path, and execution target. You can create new tasks or continue existing ones. The service reads locally saved projects from the ChatGPT or Codex desktop client and uses them as working directories. When a task runs, the service tries to open the target task in the Codex desktop app and delivers the prompt through the same channel the desktop app uses for manual input. Triggering a task is done by sending an HTTP POST request to a local endpoint. If a task has a secret key configured, you include it in the request for authentication. Task prompts support placeholder fields using double curly braces, so you can pass dynamic data in your trigger request and the service fills it in. Nested fields work with dot syntax, and any unused fields in your request get appended to the prompt as extra context. If a placeholder has no matching parameter, the request fails with a 400 error. The service can run manually with standard npm commands, binding to a local port. For permanent use, it registers as a macOS LaunchAgent that starts on login, restarts on crash, and listens on a fixed local port. Logs and task data are stored in a local data directory. The project includes a smoke test that uses fake Codex components to verify task creation, continuation, and prompt delivery without consuming actual model quota.