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

Assigning Minor Bug Fixes to Cursor

They registered maintenance tasks and minor feature fixes with Cursor every morning. The changes made by the background agent were reviewed by a developer.

Cursor

How it works

  1. They allowed the Cursor GitHub app to access the repository. This process connects the agent so it can clone the repository into a separate execution environment, run commands, or create branches.
  2. They divided maintenance, bug fixes, and minor feature additions into separate tasks. For each request, they specified the target repository and branch, allowing multiple tasks to be handled at once.
  3. They checked the results and requested further changes in each task’s conversation. While out, they sometimes registered tasks via mobile and reviewed the results later.
  4. The developer reviewed the changes and PRs created by the agent. There were times when errors occurred during new branch creation, or it was unclear whether a follow-up request was modifying an existing branch or creating a new one.

Results

It helped them handle multiple tasks at the same time. However, there were still inconveniences in processes like handling new branches.

Notes

They still developed major features or made structural changes directly in the editor.

Source

Zachary Proser 블로그 Original post ↗

Example prompt · Cursor

Please process minor maintenance tasks in Cursor as reviewable changes.

Input: repository [address], base branch [branch], task list [bugs·minor improvements], execution instructions [guidance]. First, check existing changes and repository guidelines, then organize for each task the reproduction steps, expected behavior, and completion criteria.

Separate independent tasks into different branches, and for tasks that need to modify the same file, determine the order. For bugs, first reproduce the issue, then explain the cause and fix only what’s necessary. Clearly record which branch each follow-up request applies to.

Run appropriate tests for each change and summarize the results in a reviewable PR. Do not delete failing tests or simply change expected values to make them pass. Do not write as if you ran tests that could not be executed due to environment limitations.

Deliverables: per-task changes, verification results, outstanding issues, and PRs. If branch creation fails or there are conflicts, preserve existing work and explain the cause. Do not include major structural changes or automatic merges; finish in a state I can review.