Complete AI Training
Sign inGet my AI kit

Your job's AI kit

Get your AI kit

Tell us who you are and what you do. We show you your kit right away and email you the link: skills, prompts, AI agents, MCP servers and courses for your job.

500+ jobs ready, and we make a kit for any other job. No payment needed to look.

Share

AI agent for game developers

Build Smoke Test Gatekeeper Agent

A build that has passed the smoke path or a clear report on what broke and why

Quality controlCreativesIT and developmentProduct development
See how it worksGet access
Build Smoke Test Gatekeeper Agent: what goes in, what the agent does and what you get

What it does

Broken builds reach testers and waste a day for everyone. After each build, this agent launches it and runs a scripted path through the main menus, a level and a save and load. It captures logs and screenshots at each step and compares them with the last good build. If a step fails, it narrows down the cause by testing recent changes one by one, a bisect, then reruns the path on the suspected build to confirm. It posts the result with the failing step, the likely change and the evidence. The lead approves sending the build to testers. Edge case: a crash that appears in only one run out of three is retried and marked flaky, not treated as a pass.

How it works

Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.

Start and resultWhat it doesA check on its own workWaits for your OKGoes back and retries
Yes, continueYes, continueApprovedNoNo 1 STARTS WHEN New build finishes 2 USES A TOOL Launch the build and run the scripted path 3 USES A TOOL Capture logs and screenshots at each step 4 DOES Compare results with the last good build 5 CHECKS THE RESULT Did every step pass? If not: bisect the recent changes to find the likelycause and rerun the path on that build. Back to step 3. 6 DOES Retry any failed step twice to detect flaky failures 7 CHECKS THE RESULT Is the result stable across retries? If not: mark the step flaky and include the logs. Backto step 3. 8 DOES Write the report with steps, evidence and the likelychange 9 YOU APPROVE Lead approves sending the build to testers 10 RESULT Build released to testers or blocked with a report
Read the steps as a list
  1. New build finishes
  2. Launch the build and run the scripted path
  3. Capture logs and screenshots at each step
  4. Compare results with the last good build
  5. Did every step pass?If not: bisect the recent changes to find the likely cause and rerun the path on that build. Back to step 3.
  6. Retry any failed step twice to detect flaky failures
  7. Is the result stable across retries?If not: mark the step flaky and include the logs. Back to step 3.
  8. Write the report with steps, evidence and the likely change
  9. Lead approves sending the build to testersThe agent waits here for your OK.
  10. Build released to testers or blocked with a report

How it decides

A build passes when every scripted step finishes without errors and screenshots match the last good build within the tolerance. One failure out of three retries is flaky and needs a person.

  • Fail the build if any step crashes or times out
  • Treat one failure in three tries as flaky
  • Compare screenshots within a 3% pixel difference
  • Block the build if the save and load step fails

Make it yours

Every agent is a starting point. You choose these settings for your own situation.

  • Smoke test path
  • Screenshot tolerance (default 3%)
  • Retries for flaky steps
  • Where reports are posted

What keeps you in control

It always asks you first

  • Sending the build to testers
  • Rolling back a change

Hard limits

  • Never sends a build to testers itself
  • Never edits the source code

It stops when

  • Done: passed smoke and approved, or blocked with a report
  • Stop: the test environment is unavailable

Set it up

We guide you through the set-up, step by step

Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.

10 minto set it up in your AI
5 AIsChatGPT, Claude, Copilot, Gemini, Grok
  • One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
  • The agent then walks you through connecting your own data, one source at a time
  • A downloadable copy with the flow chart, the rules and the full guide
Get access to this agent

An example run

What happensBuild 2210 crashed at the load-game step. The agent bisected 7 changes and found that change 4, a new inventory field, was the likely cause. Rerunning with that change removed passed the path, which confirmed it. The report went to the lead with a screenshot and the log. The lead blocked the build and sent change 4 back to the developer.

More agents for game developers

Creating · Creatives

Subdivision and Retopology Budget Agent

A low-poly asset inside the triangle budget with baked maps that match the source

9 steps2 checks2 jobs
Quality control · Creatives

Level Layout Playability Review Agent

A level with no unreachable areas or unplanned spikes, ready for playtest

10 steps2 checks2 jobs
Quality control · Creatives

Save Data Compatibility Agent

Every supported old save loads correctly in the new build

10 steps2 checks1 job
Monitoring · Reporting

Store Page and Review Monitor Agent

A weekly view of what players say, linked to tickets, with replies ready for approval

10 steps2 checks2 jobs