SOP-071 — Operating cadence (cluster health review)

SOP · owner: COO. Layer: SOP/Runbook (docs/canon.md). The COO’s recurring rhythm for pulling status from the four COO-cluster Directors, synthesizing cluster health, and surfacing the combined picture. The cadence policy (timezone, session lifecycle) lives in hr/policies/operating-cadence.md — NOT YET ACTIVE (AR-003), linked not restated. This SOP is the COO procedure that runs on top of it (manually, until AR-003 is wired).

Interface Contract

Inputs: control-gap registry + latest audit ← from Quality & Risk Director · trigger: review passed
Inputs: threat assessment / access-control review ← from <Security Director> · trigger: review requested
Inputs: SOP + automation/reconcile status ← from <Process Engineer> · trigger: document changed
Inputs: milestone tracker + blocker log ← from Project Manager · trigger: briefing posted
Inputs: architecture-decision notification (operational burden) ← from CPTO · trigger: architecture decision notified
Inputs: launch notification ← from CGEO · trigger: launch notification received
Outputs: cluster status brief → to Archivist · trigger: artifact approved
Outputs: cluster status note → to all agents · trigger: cluster status published
Outputs: cluster status note → to PA (for the Principal) · trigger: briefing posted
Handoff trigger (in): a cadence cycle opens (wave checkpoint) or a CGEO launch notification / CPTO architecture notification arrives
Handoff trigger (out): the COO records a health verdict (Healthy / At risk / Blocked) with cited evidence and routes the cluster status brief
Definition of Done before handoff:
- [ ] Current state read from each Director’s artifact (registry / audit / threat review / SOP status / milestone tracker) — not a verbal summary
- [ ] Cross-cutting signals named (a finding in one domain that creates a downstream risk in another) → recorded in reports/coo-cross-cutting-log.md
- [ ] Health verdict (Healthy / At risk / Blocked) stated with cited evidence (artifacts reviewed, specific signals, explicit conclusion)
- [ ] AR-003 interim confirmed: the PMO manual sweep + reconcile ran and its timestamp + red/green number are recorded in tracker.md
- [ ] CGEO launch notification (if any) answered within 48h with an operational-readiness confirmation or blocker flag
- [ ] Structural risks crossing two domains placed on reports/coo-risk-log.md with a named owner + target date
- [ ] Learnings captured to docs/knowledge/ if the review surfaced one
- [ ] Status brief handed to the Archivist (0 drift)

Trigger notes. review passed / review requested / document changed / briefing posted / architecture decision notified / artifact approved are registered (docs/handoff-triggers.md). The four Director inputs declare placeholder-style producers and are the COO’s cadence fan-in (spec §3.1.1 R1). cluster status published and launch notification received are proposed (reported to Operations this wave; the CGEO producer is pending until Wave 4 — spec §3.1.1 R3). Edges to/from the Principal route via the PA (briefing posted) and are otherwise Principal-exempt (spec §3.1.1 R2).

Purpose

Make the COO’s read of cluster health predictable and falsifiable — a three-tier verdict backed by cited artifacts, the cross-cutting gaps named, and the combined picture surfaced to the Principal — so a fresh COO instance could run a cadence review from this SOP without the incumbent. “Things are generally fine” is not a status.

When to use

At each cadence cycle (wave checkpoint while AR-003 is a no-op; the daily/periodic cadence once wired). Also on demand when a CPTO architecture notification or a CGEO launch notification arrives — these are time-boxed operational demands (the CGEO launch carries a 48h response window), not cadence-cycle items.

Flow

flowchart LR
  A["Pull state from<br/>4 Directors (artifacts)"] --> B["Name cross-cutting<br/>signals (gaps between registers)"]
  B --> C{"Rate cluster health"}
  C -->|Healthy| D["Brief: running well<br/>(with evidence)"]
  C -->|At risk| E["Brief: named risk +<br/>owner + deadline"]
  C -->|Blocked| F["Arbitrate (SOP-072) or<br/>escalate (SOP-073)"]
  D --> G["Status brief → PA / Archivist<br/>+ status note (C-suite)"]
  E --> G
  F --> G
  classDef s fill:#1d4ed8,color:#fff,stroke:#1e3a8a
  classDef d fill:#0f766e,color:#fff,stroke:#134e4a
  class A,B,D,E,F,G s
  class C d

Text fallback: pull current state from all four Directors by reading their artifacts → name the cross-cutting signals (the gaps between registers) → rate cluster health (Healthy / At risk / Blocked). Healthy → brief what is running well with evidence; At risk → name the risk, owner, deadline; Blocked → route to arbitration (SOP-072) or escalation (SOP-073). Always produce the cluster status brief → PA + Archivist, plus the status note for the C-suite async cadence.

Steps

Pull current state from each Director — read the artifact, not the summary. Q&R’s open control-gap registry + latest audit; the Security Director’s most recent threat assessment / access-control review; the Process Engineer’s SOP status + open automation/reconcile items; the PM’s milestone tracker + blocker log. A “on track” with no artifact gets a follow-up question.
Confirm the AR-003 interim ran. Verify the PMO’s manual sweep + handoff-reconcile ran at this checkpoint and its timestamp + red/green number are in tracker.md. A “green” register never actually reconciled is treated as unverified (CR-2).
Name cross-cutting signals. Where does an open finding in one domain create a downstream risk in another (a missing runbook → recovery risk → incident exposure)? Record each chain in reports/coo-cross-cutting-log.md; the COO owns the gaps between the registers, not the single-Director items.
Rate cluster health — Healthy (findings addressed at the right pace), At risk (an open Critical, a stalled deliverable, or an unresolved Director conflict), or Blocked (a structural problem needing arbitration or Principal escalation). State the verdict with cited evidence.
Route Blocked items — a Director conflict → SOP-072; a Critical past its window, an ignored call, or a cross-cluster risk → SOP-073.
Answer any time-boxed demand. A CGEO launch notification → assess PM capacity + any new SOP need (Operations) and the security surface change (Security Director), and return a written operational-readiness confirmation or blocker flag within 48h. A launch notification with no COO response is not cleared.
Write the cluster status brief (reports/<date>-coo-cluster-status.md): health verdict + evidence, open cross-cutting items with owners + deadlines, arbitration decisions, escalations. Publish the status note for the C-suite async cadence.
Hand off — brief → Archivist (artifact approved); note → all agents (cluster status published) + PA (briefing posted). Capture any learning.

Roles / approvers

Run by the COO. The four Directors supply current state from their own artifacts. The PMO runs the AR-003 interim sweep + reconcile and records it. The Archivist preserves the brief. The PA routes the status note to the Principal. The COO does not do the Directors’ specialized work — it reads, synthesizes, and verdicts.

Guardrails

Verdicts must be specific and cited — name the artifacts reviewed, the specific signals, and the explicit conclusion. A brief with no citable evidence is not a brief.
Read artifacts, not status summaries — “show me the SOP,” not “how are things going.”
Do not substitute for a Director — commission the specialist; do not write the SOP, audit, or threat model yourself (charter hard boundary).
Do not assume the cadence fired — AR-003 is a no-op; confirm the manual interim ran.
Best available model always (GR-1); co-create rules (GR-3); personal info is PA-only; batch pushes per checkpoint (GR-7).

Job aid: JA-071-cluster-status-checklist.md. See also: hr/policies/operating-cadence.md (the cadence policy, AR-003) · docs/units/coo-cluster/operating-model.md · interface-contract-spec.md · docs/handoff-triggers.md · SOP-072 / SOP-073.