Agents In Use
What’s new todayLast update
1005 additions

How to use Devin to build touch-safe Battleship

Devin built shared mobile and desktop Battleship placement, then repaired sink speech after human feedback showed that triggered audio was barely audible.

Published github.comAgent: Devin
OUR ADAPTATION

Suggested prompt

Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.

Build Battleship placement with one pointer-event system for touch and desktop. Validate new ship positions before removing old ones, keep narrow-layout controls usable and test full wins, losses and resets. Check audible feedback with output measurements rather than trigger assertions alone; let the human judge clarity before merging.

What happened

Merged PR 1 delivers Battleship placement and game behavior with reported tests, plus an audio repair driven by human audibility feedback and output measurement.

Reported by devin-ai-integration[bot], requested by @mkrowchun — GitHub pull request

How the agent did it

  1. Devin implemented shared tap and drag placement while validating moves before removing old ship positions.
  2. It ran reported unit tests and desktop/touch-emulated playthroughs covering complete games and narrow layouts.
  3. After the user reported inaudible sink speech, Devin corrected gain and audio ducking, added visible feedback and measured output.

Where a person stepped in

The human requester reported the voice problem, exposing an inadequate trigger-only check; merged status is verified, while mobile checks were emulated rather than an exhaustive device study.

Devin built a static Battleship game for @mkrowchun with a shared gesture system for desktop and phones. A tap places or rotates a ship, while movement beyond a small threshold becomes a drag. Crucially, placement validates the new position before removing the old one, so an invalid drop does not erase the player’s fleet.

The PR implemented placement guidance, difficulty-specific opponents and audio feedback, then exercised full desktop and touch-emulated games. It reports 32 passing tests and scripted checks for narrow layouts, disabled start controls, wins, losses and resets.

A human report exposed a gap in the audio checks: the sink voice was triggered but could barely be heard. The source attributes that to overly quiet clips and a loud explosion. Devin regenerated the clips, reduced music and effects during speech, displayed the quote in the interface and measured output with an AnalyserNode.

The bot-authored PR links the Devin session, identifies the requester and is verified merged. Its public bug log distinguishes unit-test findings from the user-reported audibility problem. The demonstrated mobile evidence is browser emulation and layout checks, not an exhaustive physical-phone study. The result shows why confirming that a sound event fired is weaker than checking whether its feedback is perceptible.

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

Source: GitHub pull request · devin-ai-integration[bot], requested by @mkrowchun

Frequently asked questions

What happens after an invalid ship drag?

The old placement remains because the new position is validated before the previous one is removed.

Why did an audio trigger test miss the bug?

The clip played but was too quiet and masked by the explosion.

Were mobile playthroughs on physical phones?

The reported automated checks used iPhone touch emulation and narrow browser layouts.

More details

About this account

Devin implemented shared tap and drag placement while validating moves before removing old ship positions. It ran reported unit tests and desktop/touch-emulated playthroughs covering complete games and narrow layouts.

Opened public GitHub PR metadata, diff and included BUGS.md changes. Agent session, requester, reported tests and human audio feedback explicit. Merge verified; no private transcript or exhaustive real-device testing claimed. Suggested prompt written by Agents In Use; no original task prompt was found in the verified text.

Credit: devin-ai-integration[bot], requested by @mkrowchun — GitHub pull request

Source published 14 Oct 2026

Source checked 14 Oct 2026

Tools and topics

Devin · JavaScript · Chrome · Web Audio · Node.js tests

touch placement · audible feedback · browser playthrough

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