Agents In Use
What’s new todayLast update
499 additions

How to use Claude to build an offline task app

ProgressCove’s maker used Claude Code under detailed human guidance to build a family task app with offline mobile storage. The workflow included iterative code review, tests, security review and adversarial subagent checks.

TRY A BOUNDED VERSION

Prompt to try

Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.

Build [task app] from [requirements], with optional task fields, offline mobile storage and [family account rules]. First implement the classes and components I describe in [design]. Pause for my review and apply corrections, then write tests and perform security and adversarial reviews.

What happened

The author built ProgressCove with offline mobile storage, family sharing and PIN-protected user databases; direct family login and task assignments remained planned.

Reported by twentyshores on Hacker News

How the agent did it

  1. Wrote requirements describing expected classes, attributes and objects.
  2. Had Claude Code produce an initial implementation.
  3. Reviewed the code and specified changes to refine it.

Where a person stepped in

The author reviewed and fine-tuned initial code, then requested tests, security review and adversarial review by subagents.

A person approved each step before the agent continued.

ProgressCove grew out of the author's frustration with personal task tools that imposed due dates, work-style processes or restrictions on family accounts. An earlier middleware project calculated project completion and uploaded status images for an Echo Show slideshow. The author abandoned that approach and built an app around optional fields, ongoing tasks and an absence of overdue warnings.

They used Claude Code to write code under heavy guidance. For each iteration, the author wrote requirements and described expected classes, attributes and objects. Claude produced an initial version, which the author inspected and refined by specifying changes. Afterward, they had Claude write tests, perform a security review and hand work to several subagents for an adversarial search for exploits.

The reported architecture combines Python FastAPI with a React web app and React Native mobile apps using Expo. WatermelonDB stores mobile data locally, with server synchronization and a separate PIN-protected database for each family user. Family sharing already exists, while direct family login and task assignments are described as work still to come. Who it's for: developers building personal or family task software who can supply detailed requirements and review generated code throughout implementation.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Hacker News · twentyshores

Frequently asked questions

How did the author guide Claude Code?

They wrote requirements, explained expected classes, attributes and objects, reviewed the initial code, and told Claude exactly what needed changing.

How did the app handle offline family use?

Mobile versions used WatermelonDB with synchronization. Each user had a separate PIN-protected database, selected at app launch.

More details

Remaining steps

  1. Had Claude write tests and perform security and adversarial reviews.

About this account

Claude wrote initial code and tests and performed reviews, while the author supplied detailed requirements and corrections.

Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=49894805. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.

Credit: twentyshores on Hacker News

Source published 29 Sept 2026

Source checked 14 Oct 2026

Tools and topics

Claude Code · FastAPI · React · React Native · Expo · WatermelonDB

offline first · task management · family apps

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.