Development
Everything the firm has built: Hub pages, edge functions, data flows, Project docs, artifacts, Zaps. Claude proposes a clean-up for anything stale, duplicated or exposed. Each of you makes your own call here; a clean-up is carried out only once you have both approved it, by the Sunday morning run, and the card then shows Done. Not sure? Send it back and the next session answers you on the card.
| Name | Status | Proposed | Calls |
|---|
Questions that need a partner's call. A card is Agreed when you both pick the same option, or when the named decider picks. No strong view? Press Defer to the other partner: their option then decides, and if you both defer the card goes to To discuss. Claude writes agreed outcomes up (into CLAUDE.md through the Rules tab when it is a standing rule) and closes the card. Not sure? Send it back and the next session answers you on the card. Link to a card with dev.html#d001.
Raise a new decision
What gets built next, in order. Ready to build adds a planned build or an agreed decision to the queue. On weekdays the build run builds and tests the top item in Sam's repo folder, then holds it under Waiting on Sam's push with the commit message. One at a time: nothing new starts while one waits for a push. Any Claude session can also be told "work the build queue". Already done moves an item that was finished another way (no push needed) straight to Done; Undo there puts it back.
Planned builds, not queued
Agreed decisions, not queued
Done
Every piece of automation and whether it is healthy:
About this list
Claude scheduled tasks, Supabase cron jobs and edge functions, skills (with the agents they launch), artifacts and Zaps, from the register. Cron status is read live. The rest is a snapshot that the weekday Systems inventory refresh task keeps current, and Sam's pinned Titus Scheduled Runs page shows his Claude tasks live.
How the pieces fit. Diagrams are written in Mermaid so either of you, or Claude, can edit them as text. Data flows come from the register; cron jobs and scheduled tasks are in Inventory.
Add a diagram
| ID | Flow | What moves | Status |
|---|
What the Hub's AI calls cost, by month, by source and by person, and which AI switches are on. Read only: nothing on this tab changes anything. Figures are estimates from token counts at list prices; Anthropic's invoice is the real figure.
What could go wrong with how the Hub is built or run, highest first. Close one when it is dealt with and say how.
| ID | Risk | Level |
|---|
Add a risk
Closed risks
Pending rules bind every Claude session straight away: each one reads them before it starts. The next session that works in the repo merges them into CLAUDE.md and marks them merged here.
Add a rule
Every change to the Hub's build, newest first, with how to reverse it.
| Date | Change |
|---|