HQ — the workspace above the workspaces (opt-in, and not on day one)

The problem it solves is narrow and real. Everything about a company has a home: issues, comments, runs, properties, the board. Decisions that span companies have none — which workspace a thing belongs in, whether to split one in two, a standard applied everywhere, what two of them do differently. The platform has no level above the workspace, so those decisions end up in a chat that nobody can search next month.

Create it when the second workspace appears, not before. One company needs no HQ, and an empty workspace is a slot of clutter with its own membership to maintain. This is the same twice bar the rest of the skill uses: the need is real on the second occurrence, not on the first anticipation.

Its membership is the point, not its cost. Roles are workspace-level and access is not scoped to projects (REFERENCE → the workspace is the boundary of visibility), so a separate workspace is the only way to choose who sees company-level decisions. Invite the people who make them; nobody else is looking at the wrong thing by accident.

What goes in — decisions that touch more than one workspace:

• which workspace a new product, client or experiment belongs in, and why
• splitting one workspace in two, or merging — with what moved and what did not
• a standard applied everywhere: naming, the field set, the review bar, the model tier floor
• a comparison between workspaces that led somewhere

What does not go in — anything with a home in its own workspace. A decision about one product is an issue there. Copying it up creates a second place where the truth can live, and reconciling two records is a job nobody signed up for. If you cannot say which single workspace it belongs to, it belongs here; otherwise it does not.

Referring across the boundary: <workspace-slug>/<KEY>, plus a line of substance. Issue keys are unique within a workspace — TES-1 means different things in two companies — so a cross-workspace reference carries the slug, and the sentence still reads without following it: "4teem/ACM-12 — the shared design-token decision, taken so both products stop diverging".

It does not need the methodology's field budget. HQ holds decisions, not delivery: no stages, no waves, no evidence rungs on a board that has no runs. Leave its twenty properties free, or spend two on what the owner actually filters by.

And it is not a second board. No work is executed here, no conductor runs it, no autopilot sweeps it. It is where the reasoning above the companies is written down so it can be found — and if that is not what it is being used for, it has become a fourth place to look, which is worse than none.