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

Work & ResearchWalkthrough

Agent Team That Confirms Plans and Divides Analysis of Multiple Excel Tables

TabClaw is a research agent that first confirms the analysis intent and then divides the work of multiple tables among agents.

TabClaw

How it works

  1. The author prepares CSV or Excel data to be linked together and analysis questions, then goes through a plan confirmation stage before starting the work.
  2. After confirming the intent and creating an execution plan, the author continues the analysis based on tool call results and divides the work on multiple tables among specialized agents.

Results

In the public research, the intended output is the analysis results of multiple tables and the execution plan.

Notes

This is a research prototype. The author reports checking the structure presented in the paper but did not execute it directly.

Source

TabClaw 논문 Original post ↗

Example prompt · TabClaw

Goal: Confirm Excel analysis plans and compare multiple tables.

Input & Environment: [CSV/Excel files], [criteria for linking tables], [analysis questions], [indicators to confirm].

Reference original: https://arxiv.org/abs/2606.10316

Set up and prepare connections so that a test environment can be created with the public configuration from the original paper. If there are any missing sub-workflows, APIs, or account permissions, mark them as incomplete until they are actually connected.

Connections and materials to prepare: Prepare CSV/Excel data to be linked together and analysis questions, and go through a plan confirmation stage before starting the work.

Agent actions to implement: Confirm the intent, create an execution plan, continue the analysis based on tool call results, and divide the work on multiple tables among specialized agents.

Scope of operation: This is a research prototype. The author reports checking the structure presented in the paper but did not execute it directly.

The outputs are the analysis results of multiple tables and the execution plan. Please also organize the necessary setup steps and actual execution methods. After checking the results returned by the tool, select the next task, and separately note any data that could not be accessed or tasks that could not be executed.

Verification: Input tables with different column names or duplicate join keys, confirm the plan, and compare the actual calculation results with the original rows. Clearly distinguish the inputs used for verification, the actual outputs, and any items not yet confirmed.