Ben authored commit 42110fe on June 27, rewriting OpenWork's public /start.md flow to use bootstrap-workspace instead of cloud onboarding when email verification blocks production signup.
The public thread is in the commit itself: setup pressure forced the guide to stop asking an agent to sign a human up first, create a provisional workspace and starter skill, prepare the desktop app, and let the human claim ownership later.
The human layer
Ben's GitHub profile is tied to different-ai/openwork, and his pinned public work includes agent-facing systems such as agent-bank. That context matters only because this patch is about making an agent-run setup path safe enough for a human handoff.
The code path
The patch changes the landing app route, the copy-prompt button, and the packaged bootstrap start.md. It also marks the public guide as no-store so agent crawlers do not keep stale install instructions.
Why this contribution matters
For platform readers, onboarding is not just documentation. It is a trust boundary between an agent that can prepare a workspace and a person who still has to own it.
The conversation to open
The constructive question for OpenWork is whether provisional claim flows become the default pattern for agent-prepared team software. The fairness limit: this proves one source-backed setup repair, not a finished onboarding standard.