Use an agent to triage repo work and buy groceries
They use a custom agent setup to file repo issues, carry code changes through testing, and place grocery orders under a spending limit.
- What they did
- They run oh-my-py as a personal assistant, alongside coding help. Different models handle planning, task vision, and advice, with routing through OpenRouter providers chosen from the cheapest options that still meet a minimum speed. In plain language, they can ask it to open a repo issue, then let another agent finish the code, run tests, and merge updates to master. For shopping, it pulls credentials from Bitwarden machine secrets and uses a virtual card tied to their phone.
- How it went
- They say the system completes these jobs on its own, and that work amounting to hundreds of hours can be handled from about one hour of input. No specific measurements or failures were reported.
- Tips
- Keep spending bounded with a card limit, and store credentials in a secrets manager rather than in the prompt. Tasks that barely overlap are a better fit for parallel agents.
Try it yourself with OpenClaw
Act as my personal assistant for two tasks. First, review [bug or feature request] for [repo name], write a GitHub issue draft, and stop for my approval before you create it. After I approve, hand the code changes to a separate agent, run tests, and prepare the update for master. Second, shop for [grocery items] at [store], using my saved credentials and the spending-limited virtual card on this account. Show me the issue text, the cart, and the total before any posting, submitting, checkout, payment, or deletion, and wait for my confirmation at each irreversible step.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Turn handwritten appointments and PDFs into calendar entries
A developer built a private agent that reads appointment photos, adds events, searches email, and files PDFs into paperless.
Let a sleep pose trigger a fan
They built a Raspberry Pi setup that watches sleeping posture and flips a fan on or off with a remote button pusher.
Plan meals and household tasks for the week
They used an agent to assemble weekly dinners, errands, and daily briefs around their family calendar and preferences.
Set up a private voice-controlled home assistant
They ran a local AI on their own hardware and used voice to manage home tasks without sending data to outside services.
Use a local agent for personal errands
A commenter uses a trusted local agent to handle travel, media conversion, ebook searching, design tweaks, and schedule tracking.
Scan my inbox and calendar for repeats and offer to handle them
An app built on Gmail and Calendar looks for recurring tasks, then asks whether to prep meetings, emails, or bills.