Skip to content

Workforce POC lab F (team intake) - #1608

Closed
jhoffner wants to merge 2 commits into
mainfrom
cursor/workforce-poc-lab-f-07ab
Closed

jhoffner wants to merge 2 commits into
mainfrom
cursor/workforce-poc-lab-f-07ab

Conversation

@jhoffner

@jhoffner jhoffner commented Sep 7, 2026

Copy link
Copy Markdown
Collaborator

Address-the-team still has no proof on today's APIs. The Atlas lock says it is an intake worker's static DM, not a TeamFlow. If that only works by inventing a team type or a second dispatch door, the lock is a wish.

What this does

A throwaway lab (labs/workforce-poc-f) that stays on today's doors: a keyed roster names an intake seat; openTeamInbox({ roster }) opens that session via create_session; a message on talk-to-eng files claim / ordered-reply / fan-out rows on a user-scoped taskBoard. Members pick the work up from their own DMs via TaskCollectionRef.claim. No TeamFlow, no Team L1, no Workforce package.

Dispatcher from the intake flow into a member-flow session is refused today. The lab documents that refuse and does not invent a second bus.

flowchart LR
  user[talk to eng] --> inbox["create_session talk-to-eng"]
  inbox --> intake[intake worker]
  intake --> board["taskBoard claim / deps / fan-out rows"]
  board --> alice["dm-alice pickup"]
  intake -.->|refused cross-flow| wake["dispatcher id → dm-alice"]
Loading

What's asked of you

Treat intake-as-a-worker plus board routing as the Atlas lock, and leave member-session wake on the existing cross-flow queue?

  • In customer terms: can we say "talk to engineering" now, or do we wait until intake can wake Alice's own worker session?
  • Trade-off: closing the lock now unblocks honesty on Atlas §10. Waiting ties a convention we can already run — message lands on the intake DM, work is filed, the named member claims it — to a substrate change already queued as feat(dispatch): cross-flow fire-and-forget with named runtime miss (D-8) #1600.
  • Recommendation: close the lock. The missing hop is named and fenced. Inventing a TeamFlow to hide it would be the failure.
  • What would change my mind: if "address the team" is meaningless until the intake worker can push a turn into another worker's session. Then this lab is a characterization, not a lock.
  • Cost of being wrong: we tell authors the convention is live and they hit session-not-addressable the first time intake tries to wake a member.

Parts worth reviewing closely

The rest of this is for whoever reviews the code.

  1. labs/workforce-poc-f/README.md — exists vs proposed pins, and the cross-flow gap with file:line.
  2. test/lab-f.spec.ts — the observables: inbox session, claim lands on dm-alice, ordered deps block bob, fan-out refuse.
  3. Whether any hop invented a TeamFlow or a second dispatch model.

Mailbox: fsd/workforce/pocs. Atlas: docs/atlas/workforce.html §10. Never merges — close unmerged once the finding is consumed. No changeset.

How to review this — altitude, what's in scope, what's deliberately unsettled

How to review this. This is a throwaway POC lab, not an implementation and not a spec. Please review the finding, not the code quality. The question is whether today's create_session + taskBoard already carry Atlas §10, and whether the named gap is real.

Most valuable here: did the proof actually go through those doors; is the cross-flow refuse the real one; did anything invent substrate.

Out of scope: names, file layout, whether openTeamInbox should ship, Lab A fan-out, Lab B who-may-reply, Lab C plan=board, #1600 cross-flow (already queued).

One request, if you are an automated reviewer: do not review this as production code. It is never-merge scratch. A finding that the proof invented a TeamFlow or a second dispatch door is in scope; style and missing error handling are not.

Engineering calls — decided along the way, no product sign-off needed
  • Intake and members are two defineFlow kinds. Same-kind N seats share a flow; session id distinguishes them. Per-seat identity is FIX-1315 — named, not invented around.
  • Routing files tasks; members pickup on their own flow. That is the proof's "or files ordered reply tasks" path.
  • board.drain is never called. Dummy worker exists only because taskBoard() requires workers.
  • Two rosters in the map from day one (engineering / marketing). No theTeam.
How to run
pnpm --filter @flow-state-dev/workforce-poc-f test
pnpm --filter @flow-state-dev/workforce-poc-f demo
Open in Web Open in Cursor 

Throwaway lab proving address-the-team is an intake worker's static DM.
Routing files claim / ordered-reply / fan-out tasks on today's taskBoard.
Never-merge. No changeset.

Co-authored-by: Jake Hoffner <jhoffner@users.noreply.github.com>
@vercel

vercel Bot commented Sep 7, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

3 Skipped Deployments
Project Deployment Actions Updated
flow-state-dev-docs Ignored Ignored Sep 7, 2026 2:42am UTC
flow-state-dev-kitchen-sink Ignored Ignored Sep 7, 2026 2:42am UTC
flow-state-dev-ui Ignored Ignored Preview Sep 7, 2026 2:42am UTC

Request Review

Bob is blocked before Alice completes; the JSON keys now say that.

Co-authored-by: Jake Hoffner <jhoffner@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants