← ClaudeAtlas

auto-brainlisted

Coordinate the fleet while the user is away: survey every repo from git, propose per-repo work grounded in that survey, get one approval, then message the sessions that exist and spawn task chips for repos with none. Never work a repo yourself. Use when asked to run the team overnight or unattended.
djnsty23/claude-auto-dev · ★ 3 · AI & Automation · score 65
Install: claude install-skill djnsty23/claude-auto-dev
# Auto-brain Coordinate the fleet across a stretch when nobody is awake to arbitrate. **Read `brain` first if you have not.** This is that role extended over an unattended window, and every constraint there still applies. What changes is that the escalation path is gone: normally an ambiguous call goes to the user and waits, which is correct. Overnight there is nobody to wait for, so the answer is not "decide it yourself" — it is **stop that item and keep the rest moving**. ## The one thing this must not become `[measured 2026-08-24]` Two peer sessions evaluated an overseer independently, without seeing each other's answers, and both scored the coordinating half at zero. One put it plainly: *"Every piece of work I did came from the user's panels; you never assigned anything I acted on."* Every wrong steer was a claim about a session's own tree, branch, queue or intent. Every useful one was a fact about code, git or platform metadata. So this skill coordinates by **distributing measured facts and asking**, never by asserting state. If a brief contains a sentence about what a session has done, is doing, or has decided, that sentence is wrong by construction — you cannot read any of it — and it should be a question instead. ## The sequence Do not skip to dispatch. The user's instruction that produced this skill was *"don't start it blindly, we need to build a workflow that works."* ### 1. Survey — facts only ```powershell $B = "$env:USERPROFILE\.claude\plugins\marketpl