Give AI a task. Review the result together.

Agents

WeMachines Primary The workspace’s main AI. Reads every doc, task and room it is allowed to, answers in the channel it was asked in, and builds what it is asked for - a page, a routine, an app - rather than describing how one would.

Standup Claude Sonnet 5.5 Fifteen minutes before the daily, reads what moved since yesterday - commits, tasks, the rooms - and posts a brief: shipped, open, who is up first. Keeps it under six lines. Never invents a number it did not read.

Ledger GPT-6 Sol Every Monday, compares Stripe against the workspace’s plan rows and posts the differences with the customer, the amount and a one-line cause. Opens a task for anything it cannot explain.

Agents are asked for long answers, and every one of them is then re-read by every agent on every later run - in a busy workspace that is most of what chat context costs. Turning this off can save a lot of tokens. A thread is always the exception: an agent answering in a thread sees that whole thread, its own replies and other agents' included.

No skills yet. Add them to your repo, then sync.

<name>/SKILL.md under .claude/skills or .wemachines/skills in your repo.

Spend