Umbra Group / Studio / Collaboration
← Agent Runtime v1.0 · L4
DocumentCollaboration Stack
ClassificationInternal
Versionv1.0 · 2026.04
OwnerSprint Lead
Studio Ops Kit · L4 Collaboration & Comms · Internal

Collaboration Stack.

The surface where the Studio and the client actually meet. Slack Connect channels, Gmail and calendaring, Linear for tracking, Figma for design, Loom for async demos, Google Drive for doc sharing, and the meeting cadence that holds it all together. This is Layer 4 of the stack map.

Paired with the Lighthouse Master Sprint Guide (USO-LH-02), which defines the client-facing rituals this stack supports. Owned by the Sprint Lead; read by anyone who interacts with a client. Reviewed quarterly.

§01

How to use this doc.

Audience · onboarding path · philosophy of the L4 stack

L4 is the most opinionated layer in the kit, and the one where we push back on client requests hardest. Collaboration discipline is the single biggest determinant of a sprint’s perceived quality. A brilliantly-built agent system delivered through a chaotic channel feels chaotic; an ordinary system delivered through a disciplined channel feels professional.

This doc specifies where conversations happen, where work is tracked, where demos live, and how often people meet. When a new client wants to bring their own stack (“we use Microsoft Teams”), we accommodate with a documented exception — but the cadence and discipline below hold regardless of the surface.

Reading order

  1. If you are a new Sprint Lead: read §02, §03, §08, §09 before your first client call.
  2. If you are onboarding an Agent Engineer: read §04 and §07 so they understand where Linear issues and shared docs live.
  3. If a client is asking to change a tool: re-read §03 or the relevant section and file an exception in the decision log.
§02

Collaboration at a glance.

Six surfaces · one per workload

Every collaboration need maps to exactly one surface. When two surfaces compete for the same workload, the result is fragmented history — the single most common complaint from our own post-mortems. Never put the same kind of artifact in two places.

Real-time chatPrimary

Slack Connect.

One shared channel per engagement. Fast back-and-forth, unblocking, short notices. Not for decisions.

Async + formalSecondary

Email (Gmail).

Contracts, formal notices, anything that must have a paper trail. Also where auto-alerts land for SEV-1/2.

Work trackingOwned

Linear.

Every issue with a status. All decisions land as Linear comments or linked docs. The sprint’s source-of-truth work-tracker.

Async demoUsed weekly

Loom.

Weekly demo (USP-011) recorded and shared. Also ad-hoc “how to use this” explainers for the Workflow Owner.

DesignWhen needed

Figma.

Wireframes, spec sheets, wrapper UI. One team file per engagement. View-only for clients by default.

DocsEverywhere

Google Drive.

One shared folder per engagement. Gated READMEs, deliverable docs, the state sheet. Strict folder discipline (§07).

§ Surface discipline
The fastest way to lose a sprint is to have two surfaces answering the same question. Clients often push back (“we always email for this”) — and we flex, but we document the exception and keep the per-workload surface count at one.
§03

Client channels.

Slack Connect · Gmail · Google Calendar · setup & etiquette
Channel 1
Slack Connect channel.
Primary real-time

Created on Day 1 of Discovery. Named #umbra-<client-short>-sprint. Members: Sprint Lead, Agent Engineer, Governance Architect, Executive Sponsor, Workflow Owner, Technical Counterpart.

Rules of the channel
  • No decisions here. Decisions land in Linear or a written doc; Slack summarizes and links.
  • Office hours, not 24/7. The Studio posts a pinned note stating expected-response windows (e.g., 9–18 local, M–F).
  • Threads, not walls. Top-level posts get reserved for announcements and standup summaries.
  • SEV-1 does not page via Slack. Slack mirrors the email; the page itself is the email.
Alternatives under client pressure
  • Microsoft Teams — fine if the client is already there. Same rules; slightly slower Studio response.
  • Client’s own Slack workspace (single-channel guest) — acceptable only if they can’t set up Connect.
Channel 2
Email (Gmail).
Formal + alert

Every engagement has a single canonical email thread for each deliverable hand-off (SOW, invoice, gate sign-off). Short, signed, auditable.

Rules
  • One thread per deliverable — do not mix the SOW into a support-window email thread.
  • SEV-1/2 alert emails come from a dedicated alerts@ sender, not from a human inbox.
  • All substantive emails CC the shared client alias — never single-point any client.
  • File every thread in the engagement/<client>/email/ archive at Gate 4.
Channel 3
Google Calendar.
Shared rituals

All recurring meetings live on a shared calendar attached to the engagement’s Drive folder. Recurring events are created by the Sprint Lead on Day 1 and locked against ad-hoc shuffling.

Standing events created on Day 1
  • Mon 30 min — weekly kickoff / demo plan.
  • Fri 45 min — weekly demo (USP-011).
  • Gate calls: 60 min at the end of each phase.
  • Optional: daily 15 min stand-up during Weeks 5–7 if the workload demands it.

Etiquette table — what goes where

SituationSurface
“Can you look at this real quick?”Slack Connect
“Here’s the signed SOW.”Email
“We decided to change the wrapper scope.”Linear issue + Slack link
“I recorded the weekly demo.”Loom link, posted to Slack + Linear
“Here is the gate checklist.”Linear + Drive doc, calendar invite for the gate call
“Production incident, SEV-1.”Email primary, Slack mirror
“Quarterly review reminder.”Calendar event with prep doc in Drive
§04

Project tracking.

Linear template · issue conventions · client visibility

Linear is the source-of-truth work tracker. Every sprint starts from the same template: labels, workflow states, cycle cadence, and an onboarding readme pinned in the workspace. The template exists because re-inventing the structure for each client is exactly the kind of drag we refuse.

The sprint template — 1 Linear team per engagement

ElementConvention
Team name<client-short>-sprint
Identifier prefixCLI (e.g., IDT for Indietheka)
Cycles1-week cycles, Mon–Sun, auto-rollover enabled.
Workflow statesBacklog → Todo → In progress → In review → Done → Canceled
Priority labelsUrgent, High, Medium, Low — mapped 1:1 to sprint SEV intuition.
Required labelsphase:discovery · phase:redesign · phase:build · phase:handoff · governance · agent · doc · blocker
TemplatesIssue templates for: New Agent, Governance Instrument, Pattern Candidate, Incident.
Client visibilityTeam set to visible; Workflow Owner and Technical Counterpart invited as members.

Issue conventions

  1. Every agent has exactly one Linear issue that serves as its home. Sub-issues hang off it. The agent’s spec lives linked from that issue’s description.
  2. Every governance component instrumentation has a dedicated issue with label governance — closed only when the instrumentation passes a live drill.
  3. Every decision becomes a Linear comment with a prefix DECISION: — makes the sprint decision log greppable.
  4. SEV-1/2 incidents get their own incident-labelled issue with timestamped comments.
  5. Pattern candidates land in an USP-candidate label queue reviewed at Gate 4.
§ Alternatives
If a client mandates their own tracker — Jira (common), GitHub Projects, Asana — we mirror structure (labels, cycles, states) but accept some fidelity loss. Never dual-track. One system holds the truth.
§05

Async demos & video.

Loom discipline · what gets recorded · retention policy

Async video is disproportionately valuable in a Lighthouse Sprint. Stakeholders who miss the Friday live demo can watch the recording; the Workflow Owner can rewatch a governance drill before running it themselves. Loom is the default; the discipline around it is what makes it work.

What always gets recorded

  1. Friday demo (USP-011). Always. Even if the live call had perfect attendance. Recording is the deliverable.
  2. Governance drill walkthroughs. Every drill in Week 10 (Handoff) is recorded so the Workflow Owner can review before attempting.
  3. “How to use this tool” explainers. Whenever a new tool or dashboard is handed to the client, a 3–5 min Loom replaces a meeting.
  4. Post-mortem readouts. For SEV-2+ incidents, a short Loom walks through the timeline and the five-why analysis.
  5. Sprint closeout (at Gate 4). One 10-min Loom covering what was built, what patterns were extracted, and how the support window works.

Retention & organization

RulePractice
WorkspaceOne Loom workspace for the Studio; sub-folders per client engagement.
Naming<client> · <phase> · <wNN> · <topic> — e.g., Indietheka · Build · w07 · Album-Review Demo.
Client accessShared via link with “anyone with link can view”; not public. Every link drops in Slack and in Linear.
RetentionKept through the 30-day support window. After closeout, the engagement folder is archived to Drive and wiped from Loom at client request.
Sensitive materialIf the demo shows client data or credentials on screen, the recording is redacted before sharing or re-recorded with test data.
§ Alternatives
Loom first. If a client mandates otherwise: Tella for higher-fidelity, Descript when editing/captions matter, raw mp4 to Drive for restricted environments. Zoom recordings are acceptable for the Friday demo specifically — not for everything else.
§06

Design & Figma.

Library shape · file conventions · client access

Most Lighthouse Sprints touch design lightly — a wrapper dashboard, a state-sheet visualization, occasionally a client-facing UI. We keep the Figma footprint intentionally small. One shared team file per engagement, one library file maintained by the Studio.

Studio library (shared across engagements)

LibraryContents
Umbra/CoreColor tokens, type styles (IBM Plex), eclipse mark, spectrum strip. Matches umbra-core.css.
Umbra/PatternsComponents: state-sheet mini, governance lights, severity chips, audit-row preview.
Umbra/HandoffTemplates for design-handoff specs, dashboard wireframes, wrapper layouts.

Engagement file conventions

  1. One team file per engagement — named <client> · sprint v1.
  2. Pages: Cover · Wireframes · Handoff · Archive. No stray pages; move abandoned explorations to Archive.
  3. Client access: view-only by default. Grant edit only when asked and documented.
  4. Every handoff page cross-links to the matching Linear issue.
  5. At Gate 4, the file is duplicated into the Studio archive and the live file is transferred or kept per contract.
§ Alternatives
Figma is non-negotiable for shared design work. If a client absolutely cannot use Figma: Penpot (open-source, self-hostable) is the fallback; static PNG/PDF exports with versioned filenames if all else fails.
§07

Document sharing.

Google Drive folder spec · access model · what lives where

Every engagement has one Google Drive folder. Inside, the folder structure is identical across clients. Same folders, same names, same READMEs. This is why a new Sprint Lead does not have to learn each client’s organization — we impose ours.

Canonical folder tree

FolderContentsAccess
01-discovery/Stakeholder interviews, Workflow Audit (USO-LH-08), Baseline Metrics (USO-LH-09), Discovery Report (USO-LH-15).Client + Studio
02-redesign/Redesign Blueprint (USO-LH-10), Agent Spec Sheet (USO-LH-11), allocation decisions.Client + Studio
03-build/Weekly demo recordings index, weekly build plans, incident post-mortems.Client + Studio
04-handoff/Operations Manual (USO-LH-12), Governance Runbook (USO-LH-13), training materials, Outcome Report (USO-LH-17).Client + Studio
state/The live state sheet. Clients can open and watch in real time.Client + Studio (edit for Studio)
contracts/SOW, NDA, amendments, invoices.Studio + Executive Sponsor only
archive/Everything post-Gate-4 that does not fit elsewhere. Organized by date.Studio (read-only for client on request)

Access model

  1. Client membership is issued at the folder level — never at the file level. This makes Gate-4 revocation atomic.
  2. Sensitive subfolders (contracts/) are explicitly exceptions — not inherited.
  3. No external sharing of any file in these folders beyond the named members without a written approval trail.
  4. At Gate 4, client membership is moved to view-only on everything except state/ (transferred if contract specifies).
  5. At Day 30 post-Gate-4, Studio access reviews: any sensitive credentials in the folder must be revoked if they were shared during delivery.
§ Alternatives
Google Drive is the default. If a client lives in Notion or SharePoint, we mirror the folder tree there, maintain a canonical Drive copy, and flag the duplication cost in the §04 cost model. Never abandon the canonical copy.
§08

Meeting cadence.

What recurs · what does not · why

A Lighthouse Sprint has five recurring meetings and four gate calls — no more. We aggressively resist expansion. Meetings that were added mid-sprint and never removed are the silent killers of engagement satisfaction.

MeetingCadenceDurationOwnerOutput
Weekly kickoffMonday30 minSprint LeadWeek’s plan; Linear cycle scoped.
Weekly demo (USP-011)Friday45 minSprint Lead + Agent EngineerLoom recording; audience feedback captured in Linear.
Daily stand-up (optional, Weeks 5–7 only)Daily15 minSprint LeadUnblocks; no artifacts.
Gate callEnd of each phase (4 total)60 minSprint LeadSigned gate checklist; next-phase kickoff.
Incident reviewAs needed (SEV-1/2 only)30–60 minGovernance ArchitectPost-mortem doc.

Meeting discipline

  1. Every recurring meeting has a pinned prep doc. If the prep doc is empty at T-4 hours, the meeting is canceled — not postponed.
  2. No meeting exceeds 60 minutes. Long syncs are a sign that async is failing somewhere; find it.
  3. No standing one-on-ones with the client. The shared channels are where conversations happen; bilateral back-channels create fragmented history.
  4. Every meeting has a written output posted to Linear or Drive within 24 hours — not just a calendar history.
  5. Gate calls always happen. Even if the gate is a rubber-stamp pass. Skipping erodes the discipline for the next one that matters.
§ The rule
Every new recurring meeting proposed must retire an existing one. We hold this line aggressively. Sprints that accumulate meetings accumulate ceremony without accumulating progress.
§09

Client onboarding.

Day-0 setup · Day-1 ready · what the client sees

Onboarding happens in the 48 hours between SOW countersignature and Day 1 kickoff. The goal is that when the client shows up to the kickoff, every surface they need is already live and waiting.

Day-0 (T-48 h) — Studio-side setup

  1. Create Google Drive folder with canonical tree (§07). Pre-populate readme.md in each subfolder.
  2. Create Linear team with sprint template (§04). Pre-create the four gate issues.
  3. Create shared Google Calendar; schedule all recurring events for the 11-week sprint + 30-day support window.
  4. Create Figma team file with Studio library applied.
  5. Create Loom folder for the engagement.
  6. Draft the Slack Connect channel; send invite to the client’s designated Executive Sponsor.
  7. Pre-fill state sheet template; link it inside state/.

Day-0 (T-24 h) — welcome email

One email to the Executive Sponsor and named client team, with:

  1. Drive folder link — one click, which reveals the full canonical tree.
  2. Slack Connect invite link.
  3. Linear membership invite.
  4. Calendar invites for Week 1’s kickoff and Friday demo.
  5. One short Loom (< 4 min) walking through what the sprint will look like in the first two weeks.
  6. Contact escalation path: who emails whom for SEV-1 in the off-chance it happens in the first week.

Day-1 kickoff — 60-min agenda

  1. Introductions (5 min).
  2. Walk through the Drive folder and what each subfolder will hold (10 min).
  3. Walk through the Linear board and the gate structure (10 min).
  4. Confirm meeting cadence; adjust recurring times if necessary (5 min).
  5. Kick off Discovery: first stakeholder interview scheduled, Workflow Audit opened (20 min).
  6. Closing: what the Studio needs from the client in the next 48 h (10 min).
§ Onboarding bar
If the Day-1 kickoff feels ceremonial, onboarding worked. If it feels like scrambling — “where’s the folder,” “can you send the invite,” “when’s the next call” — we failed before starting. The kickoff should be the first real work, not the setup.
§10

Sprint Lead checklist.

Per-engagement · per-week · per-gate

Per engagement (Day 0 → Gate 4)

  1. Drive folder, Linear team, Loom folder, Figma file, Slack Connect channel, Google Calendar all live by T-24 h.
  2. Welcome email sent with all six surfaces linked.
  3. At Gate 4: folder access downgraded to view-only; Linear team archived; Slack Connect closed per Security §05.

Per week (every Friday)

  1. Demo recorded, named per convention, dropped in Slack and Linear.
  2. Cycle rolled in Linear; incomplete items triaged (complete / carry-over / cancel).
  3. Next week’s prep doc drafted before leaving for the weekend.
  4. Slack channel housekept — unresolved threads chased or escalated.

Per gate (at phase boundaries)

  1. Gate issue in Linear has all criteria checked off with linked evidence.
  2. Gate call scheduled with prep doc pinned.
  3. Gate decision (pass / conditional / hold) logged in decision log.
  4. Calendar invites for the next phase’s recurring events reviewed and adjusted if needed.
§ Bottom line
A Sprint Lead who runs this stack cleanly will feel two moves ahead of every client. That “feels organized” quality is not a personality trait — it is the L4 discipline in this doc, executed consistently. It is the single highest-leverage thing a Sprint Lead does.