Waiting around while OpenHands clone repos, install dependencies, and prepare test harnesses?
I put together a small repo showing how custom OpenHands sandbox images can cut setup time for heavyweight repos:
https://github.com/rajshah4/openhands-custom-image
The example uses a VS Code benchmark where the agent fixes a real bug and validates it with the repo’s test harness. The custom image prebakes the expensive setup: repo checkout, node_modules, transpiled output, Electron artifacts, native packages, and verification wrappers.
My motivation was that customers have asked for this capability, and I wanted to see how much improvement is possible. I found out it can save a lot of time. The benchmark was built for Enterprise OpenHands, but the same benchmark also shows improvement with custom images for Agent Server locally.