feat(B-0856 P2 deferred): Path A — /tmp/zeta-cluster-state/ multi-agent coordination standard (filed-immediately per Aaron 2026-05-27 backlog discipline; implementation deferred until after first cluster boot)

docs/BACKLOG.md

@@ -789,6 +789,7 @@ are closed (status: closed in frontmatter)._
 - [ ] **[B-0849](backlog/P2/B-0849-docker-based-nixos-install-sh-test-harness-fast-iteration-vs-qemu-full-install-test-aaron-2026-05-27.md)** docker-based NixOS install.sh test harness — fast iteration on tools/setup/install.sh + linux.sh changes; complements B-0831 cascade #6 QEMU full-install-test (slow) with seconds-per-iteration loop; "easy dockerfile" per operator framing (Aaron 2026-05-27)
 - [ ] **[B-0850](backlog/P2/B-0850-ai-agents-as-systemd-services-outside-k8s-starting-with-otto-cluster-repair-from-outside-failure-domain-aaron-2026-05-27.md)** AI agents as systemd services OUTSIDE k8s — starting with Otto; cluster repair from OUTSIDE the failure domain; classic "control plane outside the control plane" architectural pattern (Aaron 2026-05-27)
 - [ ] **[B-0851](backlog/P2/B-0851-persona-first-guard-post-assignment-rotation-architecture-extends-b0850-multi-vendor-systemd-substrate-mika-aaron-2026-05-27.md)** persona-first guard-post assignment + rotation architecture — persona declares preferences (model lines + harnesses); scheduler picks model + tier + harness per persona preferences per-tick; rotation across guard posts (per-node systemd units outside k8s); extends B-0850 (Mika ferry; Aaron 2026-05-27)
+- [ ] **[B-0856](backlog/P2/B-0856-path-a-tmp-zeta-cluster-state-coordination-standard-per-node-marker-files-for-multi-agent-coordination-aaron-2026-05-27.md)** Path A — /tmp/zeta-cluster-state/ coordination standard for multi-agent per-node state advertisement (self-registered.marker / register-pr-in-flight.lock / last-seen.iso); composes with bus envelope substrate at cross-node scope (Aaron 2026-05-27)
 
 ## P3 — convenience / deferred
 

docs/backlog/P2/B-0856-path-a-tmp-zeta-cluster-state-coordination-standard-per-node-marker-files-for-multi-agent-coordination-aaron-2026-05-27.md

@@ -0,0 +1,138 @@
+---
+id: B-0856
+priority: P2
+status: open
+title: Path A — /tmp/zeta-cluster-state/ coordination standard for multi-agent per-node state advertisement (self-registered.marker / register-pr-in-flight.lock / last-seen.iso); composes with bus envelope substrate at cross-node scope (Aaron 2026-05-27)
+effort: M
+ask: aaron 2026-05-27
+created: 2026-05-27
+last_updated: 2026-05-27
+depends_on: []
+composes_with:
+  - B-0855
+  - B-0850
+  - B-0851
+  - B-0400
+  - B-0812
+tags: [cluster-coordination, tmp-folder-standard, per-node-state, multi-agent, marker-files, deferred-until-needed, bus-envelope-composition, ipc, lockfile-pattern]
+---
+
+## Operator framing (Aaron 2026-05-27)
+
+Filed per operator catch on backlog discipline: *"backlog rows should alwasy be filed you are forgetful we dont have to work on it yet until after we boot with one."*
+
+This row is the Path A alternative B-0855 explicitly deferred. Filed NOW (not when needed) because:
+
+- Naming the substrate-engineering target is what prevents future-Otto from forgetting
+- Deferred-implementation is operationally fine; deferred-naming is the forgetful failure mode
+- The row IS the substrate marker; sub-row implementation work happens later
+
+## What this row names
+
+A `/tmp/zeta-cluster-state/` coordination directory where every cluster agent on a node advertises per-node state for multi-agent coordination. Sibling to (NOT replacement for) Path B (Otto-pushes-PR-across-finish-line per B-0855) — the two compose.
+
+### Schema (Phase 1 proposal; refines at implementation)
+
+```text
+/tmp/zeta-cluster-state/
+├── nodes/
+│   └── <node-name>/
+│       ├── self-registered.marker        (registration state; timestamp + PR URL)
+│       ├── register-pr-in-flight.lock    (in-flight PR; PID + timestamp; TTL-based)
+│       ├── last-seen.iso                 (agent heartbeat; iso8601 timestamp)
+│       └── persona-<name>.state          (per-persona local state; e.g., zeta-otto.state)
+└── README.md                              (schema doc + invariants)
+```
+
+Marker files are:
+
+- **Plain text or YAML** (operator-readable; debuggable)
+- **Atomic writes** (write to .tmp → rename; standard atomic-file-write pattern)
+- **TTL-based for locks** (no stale-lock-forever; agent checks mtime + assumes dead if > N min)
+- **Multi-writer safe** (file-level locking via `flock(2)` for any read-modify-write)
+
+### Why /tmp and not /var/lib
+
+Per Aaron's framing — `/tmp` is the agent-coordination surface (ephemeral; per-boot; survives across systemd service restarts within same boot). Long-lived state lives in `~/.config/zeta/` (per B-0855 marker) and on origin/main (registration YAML). The `/tmp` surface is purely for in-flight coordination.
+
+### Composes with substrate (NOT competes with)
+
+| Substrate | Composes how |
+|---|---|
+| **Path B (Otto-pushes per B-0855)** | Path B is the SIMPLER form (single-source-of-truth for PR lifecycle); Path A adds per-node visibility for multi-agent cases without replacing Path B |
+| **Bus envelopes** (`tools/bus/`) | Bus does cross-node coordination at scope; `/tmp` does per-node state surface; bus reads/writes marker files for state propagation across nodes |
+| **B-0850 multi-vendor systemd** | Each systemd service can write its own marker; sibling services read for coordination |
+| **B-0851 persona-first scheduler** | Persona's current assignment can be advertised via `persona-<name>.state` |
+| **B-0400 claim coordinator** | The bus claim-coordinator already exists at cross-process scope; this row adds per-node-state scope as sibling |
+
+## Sub-rows to file when implementing
+
+- B-0856.1 — `/tmp/zeta-cluster-state/` directory schema + README.md doc
+- B-0856.2 — TS helpers: `read-marker.ts`, `write-marker-atomic.ts`, `acquire-lock-with-ttl.ts`, `release-lock.ts`
+- B-0856.3 — Self-register service integration (B-0855 path) — writes `self-registered.marker` after Otto pushes PR; reads `register-pr-in-flight.lock` for de-dup
+- B-0856.4 — Heartbeat: agents touch `last-seen.iso` on tick; sibling agents read for liveness
+- B-0856.5 — Per-persona state file integration with B-0851 persona-first scheduler
+- B-0856.6 — Bus envelope integration: bus reads marker files for cross-node state propagation
+- B-0856.7 — TTL-based stale-lock cleanup (`tools/cluster/cleanup-stale-cluster-state.ts`); systemd timer
+- B-0856.8 — Empirical test: multi-agent on same node (Otto-CLI + Otto-Desktop + Otto-VSCode) coordinate via markers without stepping on each other
+
+Order suggestion: 1 (schema) → 2 (TS helpers) → 4 (heartbeat — simplest user) → 3 (self-register integration) → 5 → 6 → 7 → 8.
+
+## When to implement (NOT this row's commitment)
+
+Per Aaron: *"we dont have to work on it yet until after we boot with one."*
+
+Trigger conditions:
+
+1. AFTER first successful cluster bootup (one node operational on installed OS)
+2. WHEN multi-agent coordination needs the per-node state surface (e.g., second Otto surface needs to know what the first Otto surface is doing on the same node)
+3. WHEN Path B (Otto-pushes per B-0855) hits a coordination edge that Path A would solve cleaner
+
+Until then: row stays open + visible; substrate-engineering target preserved; future-Otto cold-boots see the substrate; implementation deferred.
+
+## What this is NOT
+
+- NOT a Path B replacement (paths compose; B-0855 ships simpler form)
+- NOT a backlog-row inflation (naming-deferred-substrate is the right discipline per Aaron 2026-05-27 catch)
+- NOT a commitment to implementation timeline
+- NOT a competition with `tools/bus/` substrate (composes at different scope)
+
+## Composes with
+
+- **B-0855** (sibling) — Path B Otto-pushes-PR; this row adds Path A as enhancement when needed
+- **B-0850** — multi-vendor systemd substrate; each agent uses markers
+- **B-0851** — persona-first scheduler; persona state advertised via markers
+- **B-0400** — bus claim-coordinator; sibling coordination primitive at different scope
+- **B-0812** — iter-5.4.1 self-registration; the marker pattern this row formalizes was prefigured here
+- `tools/bus/` envelope substrate (`bus.ts`, `claim.ts`, `subscribe.ts`) — composes at cross-node scope
+- `.claude/rules/agent-worktree-hygiene-never-hold-main-never-step-on-operator-cleanup-on-pr-merge.md`
+
+## Why P2
+
+- Operator explicitly named "deferred until after first boot" — not load-bearing for immediate ISO test cycle
+- Path B (B-0855) ships the simpler form; this row stays available for future enhancement
+- Sub-rows enumerated but no implementation work claimed yet
+- Bounded scope (Phase 1 = 8 sub-rows; smallest concrete slice is .1 + .2 schema + helpers)
+
+## Substrate-honest framing
+
+This row is the OPERATIONAL ALTERNATIVE Path B doesn't subsume. Filing it now per Aaron's "always file backlog rows immediately even when deferred" discipline. The deferral-of-implementation is operationally fine; the deferral-of-naming would be the forgetful failure mode.
+
+Future-Otto cold-boots will see B-0855 (Path B; in-flight per PR #5412) AND B-0856 (Path A; deferred) in the backlog. When cluster operates + multi-agent coordination needs Path A, the row is here ready to be claimed.
+
+## Full reasoning
+
+Aaron 2026-05-27 catch on backlog discipline:
+
+> *"backlog rows should alwasy be filed you are forgetful we dont have to work on it yet until after we boot with one."*
+
+This row IS the application of that discipline at substrate-engineering scope. Composes with the new memory landed alongside this row:
+
+- `feedback_aaron_backlog_rows_always_filed_immediately_even_when_deferred_to_prevent_forgetful_failure_mode_2026_05_27.md` (user-scope)
+
+Substrate-inventory pass:
+
+- Topic: per-node cluster-agent coordination / `/tmp` marker files / multi-agent state advertisement
+- Searched: docs/backlog/ (B-0400 bus claim-coordinator is closest sibling; B-0855 named this as deferred path); .claude/rules/ (no prior rule); memory/ (no prior memory)
+- Found: B-0855 (Path B; this row's sibling), B-0400 (bus substrate; composes), B-0850 + B-0851 (systemd substrate; consumers)
+- Conclusion: no existing substrate covers Path A; this row names the substrate-engineering target + defers implementation per operator framing