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

Work & Research

Explaining the Processing Flow When Sending a Message to Grok Bot

This case guides users step by step on how the server and agent process a message when it is sent to Grok Bot.

How it works

  1. The user enters a message in the app and presses the send button.
  2. The app immediately displays the message and saves it locally with a unique ID.
  3. When the message is sent to the server, the server filters out duplicate messages and adds it to a queue.
  4. Each conversation has its own queue, and messages sent directly by the user interrupt the agent's current task and are prioritized.
  5. The server uses a workflow engine called Temporal to save the result of each step, and if there is a server failure, processing resumes from where it left off.
  6. When the agent receives the message, it handles simple tasks directly, while complex or time-consuming tasks are delegated to a separate helper workflow.
  7. Results of tasks or informational messages are delivered to the user through a tool called SendToUser.
  8. Shell commands and other actions executed during tasks are pre-reviewed by a separate model to prevent dangerous actions.
  9. Conversation history, files, and skills are safely backed up on the server and in object storage.
  10. Even if multiple Grok Bots are created, they share a single VM (virtual machine), and each bot has its own desktop environment.

Results

When a message is sent to Grok Bot, the app displays it immediately, and the server prevents duplicates by adding it to a queue. The agent processes tasks based on message priority, and even in the event of interruptions or server failures, previous tasks are not repeated. Simple tasks are handled directly, while complex ones are delegated to a helper, and all results are delivered to the user through a single tool. Conversations, files, and skills are safely stored on the server and in object storage, and multiple bots share one VM.

Notes

Some details about the server's internal structure and workflow implementation are based on assumptions, and the actual message processing flow and data backup methods may differ according to Grok Bot's official announcements. The author states that this explanation is based on Lee Robinson's presentation and their own experience.

Source

peterp.me Original post ↗

Example prompt · Grok Bot

Goal: Check step by step how the server and agent actually process a message when you send it to Grok Bot.

Input: Message to send to Grok Bot (e.g., 'Tell me today's schedule'), Grok Bot account and access environment (web, mobile, Slack, etc.)

Environment: Account with Grok Bot activated, client capable of sending messages (app, web, Slack, etc.), connected to the server

Implementation and setup steps:
1. Enter and send a message to Grok Bot.
2. Check if the app displays the message immediately.
3. Check if the server adds the message to the queue without duplication.
4. Observe whether multiple messages sent quickly are processed together or individually.
5. Check if the message interrupts the agent's current task or is simply added to the queue.
6. Check if the server uses a workflow engine like Temporal to save the result of each step and resumes from the interrupted part in case of a failure.
7. Check if the agent handles simple tasks (e.g., adding a memo, editing a comment) directly, and delegates complex or time-consuming tasks (e.g., processing large files, integrating with external systems) to a helper workflow.
8. Check if the result of the task is delivered to the user through the SendToUser tool.
9. Check if potentially dangerous commands (e.g., shell commands) are pre-reviewed by a separate model before execution, and if tasks requiring approval return as drafts.
10. Check if conversation history, files, and skills are backed up on the server and in object storage, and if multiple Grok Bots share a single VM.

Deliverable: Step-by-step confirmation of message transmission, app and server processing flow, queue management, task delegation, result delivery, and data backup and sharing structure

Verification criteria: For each step, confirm actual operation using the app screen, server logs, result messages, etc., and also test exception cases such as interruption, restart, and handling of dangerous commands.