AI Agent Use Cases
Views —

AI Agent Use Cases

Added today

44All 189

Prompt to Prepare a Shopping Cart from a Recipe Photo

The author reports that Grok Bot receives a recipe photo, organizes the ingredients and quantities, and creates a shopping list by excluding ingredients already available at home.

Grok Bot
Life & Hobbies

1–20 of 189

Dev

Cloud Bot Takes Over Tasks Uploaded by Local Computer Agent

The author set up a system where, when a local agent uploads a task document, GitHub Actions calls the Grok Bot webhook so the bot can start working.

Grok Bot

How it works

  1. The author created separate Grok Bots for each function and used the webhook feature to trigger them with external events.
  2. When the local agent writes up a task as a document and uploads it to the grokbot-workspace repository, GitHub Actions calls the bot's webhook. The bot reads the synced document and starts the task.
  3. The author connected Pi Agent and DeepSeek to Grok Bot to worry less about X subscription limits, and also enabled the use of models from the GPT subscription.

Results

The author reports that Grok Bot became a lightweight cloud task assignment system for the local agent.

Notes

There are no workflow configuration files or actual processed task examples in the article.

Source

X Original post ↗

Example prompt · Grok Bot

When your computer's coding agent uploads a task document to a repository, set it up so a cloud bot reads the document and starts working.

Input: Repository [name] for uploading task documents, task document folder and format [e.g., markdown files in the tasks folder], bot to trigger [name], list of tasks the bot is allowed to do. First, check if this bot can create a webhook that runs on external requests, and find out its URL and authentication method. Only put secret keys in the repository's secret settings—never leave them in files or logs.

Create a GitHub Actions workflow that only runs when there are changes in the task document folder, and make it call the bot's webhook with the path of the changed document. On the bot side, have it read the document at the received path and keep a record so it doesn't process the same document again. If a task isn't described in the document or is outside the bot's allowed scope, have the bot ask you before proceeding.

When saving results back to the repository, only write to a separate branch or results folder—never merge directly into the main branch. Also describe how to disable the workflow.

Deliverables: workflow file, bot configuration summary, sample task document, processing log format. Upload a short test document and check that the webhook is called only once and the bot leaves a result before using it for real tasks.