feat(fleet): config-type presets + AI-free init wizard

Add four config-type preset rosters (general, coding, research, hybrid) with
exactly one orchestrator each using the claude runtime; worker agents use pi
with model_hint openai-codex/gpt-5.5:high. Extend `mosaic fleet init` with a
readline-based interactive wizard (AI-free) that prompts for preset selection
when no --profile is given and stdin is a TTY; falls back to "general" with a
note when non-interactive. Export parseInitProfile, resolvePresetFilename,
countOrchestrators, FLEET_PROFILES, and FleetProfile for testability. Add 24
new tests covering preset shape, sanitization, wizard TTY/non-TTY paths, and
R8 idempotency. All gates green: test (428/428), typecheck, lint, prettier,
and the sanitization gate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RMoEx7hfdFGjUiCHuN1RRi

docs/fleet/PRD.md

@@ -0,0 +1,109 @@
+# PRD — Fleet Phase 2: Operator Observability
+
+> **Workstream:** W-FLEET under `mvp-20260312` · **Phase:** 2
+> **North star:** [docs/fleet/north-star.md](./north-star.md)
+> **Source umbrella PRD:** [docs/PRD.md](../PRD.md) (Mosaic Stack v0.1.0)
+> **Tracks task:** `fleet-observability-1` — restore operator observability into fleet agent sessions.
+
+## Problem
+
+The durable tmux fleet runs on the isolated `mosaic-factory` socket. That isolation
+(which protects the operator's default tmux) makes the fleet **invisible** to default
+tooling, and truth is split across three planes no single command joins — systemd
+(`systemctl --user`), tmux (`-L mosaic-factory`), and the process tree (`pstree`).
+`agent tail` (`capture-pane`) returns **blank for full-screen TUIs**, and `agent send`
+confirms only keystroke injection, not acceptance. Net: the operator has near-zero
+observability and no safe way to watch a session.
+
+## Goals
+
+1. One command shows the **whole fleet's** real state, joining all three planes.
+2. **Liveness is truthful**: healthy = answered a heartbeat, not "pane alive".
+3. The operator can **watch** any session read-only without disrupting it.
+4. `send` reports **delivered-and-accepted**, not just injected.
+5. Every record/address carries **`tenant_id` + `host`** (zero foreclosure for multi-tenant/multi-host).
+
+## Non-goals (this phase)
+
+- No webUI (Phase 5; rides federation for cross-host).
+- No `fleetd` daemon or persistent history store.
+- No real-runtime swap (Phase 3) — instrument the live **dogfood stub** fleet.
+- No cross-host aggregation yet (addressing is host-tagged but queries stay local).
+
+## Functional requirements
+
+| ID   | Requirement                                                                                                                                                                                                                                                                                                    |
+| ---- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| FR-1 | `mosaic fleet ps [--json]` prints one row per roster agent joining: name · tenant · host · runtime · systemd(active/enabled) · pane(alive/dead) · pid · idle · **last-heartbeat age** · **drift** flag (roster runtime ≠ actual pane command) · **boot-enable** warning (active but `UnitFileState=disabled`). |
+| FR-2 | **Heartbeat protocol v1** (see below); `dogfood-agent.py` implements the responder. `fleet ps` issues probes (or reads last-seen) and reports health per FR-1.                                                                                                                                                 |
+| FR-3 | `mosaic agent watch <name>` opens a **read-only** view of the pane (grouped session or `tmux attach -r`) that cannot send keystrokes and does not shrink the agent's window.                                                                                                                                   |
+| FR-4 | `mosaic agent attach <name>` remains the **explicit** interactive-takeover path (separate verb, documented as the only one that can type).                                                                                                                                                                     |
+| FR-5 | `mosaic agent send <name> --verify` confirms the message was **accepted** (not left as an unsubmitted draft) and returns non-zero if delivery cannot be verified.                                                                                                                                              |
+| FR-6 | All structured output (`--json`) includes `tenant_id` and `host` fields.                                                                                                                                                                                                                                       |
+
+## Heartbeat protocol v1
+
+- **Probe:** operator/`fleet ps` writes a sentinel line to the agent's input or a
+  well-known per-agent heartbeat file path `~/.config/mosaic/fleet/run/<agent>.hb`.
+- **Response:** the runtime updates `<agent>.hb` with `ts=<iso8601> pid=<pid> status=<ok|busy>`
+  on a fixed interval (default 15s) and on demand when probed.
+- **Health rule:** `healthy` if `now - ts <= 3 × interval`; else `stale`; missing file = `unknown`.
+- **Contract:** every runtime (dogfood stub now; claude/codex/pi/opencode in Phase 3)
+  MUST emit the heartbeat. The protocol is file-based so it works for headless stubs and
+  full-screen TUIs alike (no `capture-pane` dependency).
+- `ASSUMPTION:` file-based heartbeat (vs in-pane echo) — chosen because it is TUI-safe and
+  uid-scoped, fitting per-tenant isolation. Open to an OTEL-span variant in Phase 3 (MVP-X6).
+
+## Acceptance criteria
+
+- `mosaic fleet ps` shows all 5 live sessions on `mosaic-factory` with correct
+  pane/pid/idle and flags the dogfood **drift** (`canary-pi` runtime=pi but pane runs
+  `dogfood-agent.py`) and the **boot-enable** gap (active but disabled).
+- Killing one agent's pane flips its row to dead/stale within one `interval`.
+- `agent watch` shows live output and provably cannot type into the pane; detaching
+  leaves the agent's window size unchanged.
+- `agent send --verify` returns success on an accepting pane and non-zero on a wedged/draft pane.
+- Quality gates green: `pnpm typecheck`, `pnpm lint`, `pnpm format:check`, plus
+  `pnpm --filter @mosaicstack/mosaic test`.
+- Independent review passed; dogfood evidence captured against the live fleet.
+
+## Test plan
+
+- Unit/CLI specs in `packages/mosaic/src/commands/fleet.spec.ts` (and a new
+  `fleet-ps`/`watch`/`send-verify` spec) using the injected `CommandRunner` to assert
+  exact tmux/systemd command construction and JSON shape (tenant+host present).
+- Situational: run against the live `mosaic-factory` fleet; capture `fleet ps` output,
+  a kill-and-detect cycle, a read-only `watch`, and a `send --verify` pass/fail pair.
+
+## Known limitations
+
+- **Verify heuristic is best-effort:** `agent send --verify` uses a `>` -prefix draft
+  heuristic that is specific to pi/claude TUIs. Draft detection for codex and opencode
+  TUIs is best-effort only; those runtimes may not use the same input-line indicator.
+- **Pane-change check is the best Phase-2 signal; verify now polls up to a bounded
+  timeout:** `agent send --verify` captures a BEFORE snapshot, sends the message, then
+  polls `capture-pane` every ~400 ms up to a configurable total timeout (default ~6 s,
+  controlled by `--verify-timeout <ms>`). On each poll it runs classifySendResult: if
+  the pane shows 'accepted' or 'draft' the loop exits immediately; while the result is
+  'unverifiable' (no pane change yet) it keeps polling. After the timeout with no
+  definitive result, it fails closed: exit 1 with "no pane change after send". This
+  eliminates false 'unverifiable' failures for slow/loaded TUIs that were previously
+  caused by the old fixed 300 ms single-capture. Definitive acceptance ultimately
+  requires a runtime acknowledgement (Phase-3 heartbeat-ack); the bounded pane-change
+  poll is the best signal available against an opaque TUI for Phase-2.
+- **Blank AFTER capture fails closed:** Full-screen TUIs (claude, codex, opencode, pi)
+  render blank for `tmux capture-pane`. When the AFTER snapshot is empty, `send --verify`
+  returns non-zero with an "unverifiable" message rather than silently succeeding. This
+  is an intentional fail-closed design (FR-5).
+- **`agent watch` uses a grouped viewer session:** `tmux attach -r` directly against the
+  agent session lets the viewer terminal shrink the agent's window. `agent watch` instead
+  creates a throwaway grouped session (`tmux new-session -d -t '=<agent>' -s
+'<agent>-watch-<pid>'`), attaches read-only to that session, and kills it on detach.
+  The grouped session shares the agent's windows but has independent sizing, so the
+  agent's window is never affected. `tmux attach` is still interactive and requires
+  inherited stdio; the `interactiveRunner` handles TTY passthrough.
+
+## Surfaces & parity (MVP-X1)
+
+CLI lands this phase. TUI surface follows in the `packages/mosaic` wizard; webUI in
+Phase 5 via federation. PRD records the parity debt explicitly so it is not lost.

docs/fleet/TASKS.md

@@ -0,0 +1,27 @@
+# Tasks — W-FLEET (Fleet) Phase 2: Observability
+
+> Workstream task file for the Fleet. Single-writer: Fleet workstream lead (orchestrator).
+> Workers read but never modify. This is **not** the MVP rollup (`docs/TASKS.md`) — a
+> rollup row is proposed to the MVP orchestrator, not written here.
+>
+> Mission: `mvp-20260312` · PRD: [docs/fleet/PRD.md](./PRD.md) · North star: [docs/fleet/north-star.md](./north-star.md)
+> Status: `not-started` | `in-progress` | `done` | `blocked` | `failed`
+
+| id            | status      | description                                                                                                        | depends_on            | agent       | pr  | notes                                                                                                                         |
+| ------------- | ----------- | ------------------------------------------------------------------------------------------------------------------ | --------------------- | ----------- | --- | ----------------------------------------------------------------------------------------------------------------------------- |
+| FLEET-OBS-000 | done        | Plan: north-star + Phase-2 PRD + workstream scaffolding                                                            | —                     | lead        | —   | persisted 2026-06-20 on `feat/fleet-observability`                                                                            |
+| FLEET-OBS-001 | done        | Heartbeat protocol v1 spec finalized in PRD + framework doc                                                        | FLEET-OBS-000         | lead        | —   | file-based `~/.config/mosaic/fleet/run/<agent>.hb`; spec in PRD                                                               |
+| FLEET-OBS-002 | in-progress | Implement heartbeat responder in `dogfood-agent.py`                                                                | FLEET-OBS-001         | fleet-coder | —   | dispatched to ad-hoc `mosaic yolo` fleet agent (dogfood)                                                                      |
+| FLEET-OBS-003 | done        | `mosaic fleet ps` — join systemd+tmux+proc+idle+heartbeat; tenant+host tagged; drift + boot-enable flags; `--json` | FLEET-OBS-001         | worker      | —   | commit ab47831; LIVE-verified on mosaic-factory; caught canary-pi DRIFT + BOOT-ENABLE. Polish: idleSeconds parse returns null |
+| FLEET-OBS-004 | done        | `mosaic agent watch <name>` — read-only join (no resize, no keystrokes)                                            | FLEET-OBS-000         | worker      | —   | `attach -r`; verb wired                                                                                                       |
+| FLEET-OBS-005 | done        | `mosaic agent send --verify` — delivery/acceptance receipt                                                         | FLEET-OBS-000         | worker      | —   | --verify flag; draft-heuristic verify                                                                                         |
+| FLEET-OBS-006 | done        | CLI specs for ps/watch/send-verify (tenant+host shape, command construction)                                       | FLEET-OBS-003,004,005 | worker      | —   | 62 tests green (31 new); re-verified by lead                                                                                  |
+| FLEET-OBS-007 | not-started | Framework doc: fleet observability guide + verbs                                                                   | FLEET-OBS-003,004,005 | lead        | —   | `docs/guides/` or `framework/tools/.../README`                                                                                |
+| FLEET-OBS-008 | not-started | Independent review + dogfood verification on live fleet                                                            | FLEET-OBS-002..007    | reviewer    | —   | author ≠ reviewer; capture evidence in scratchpad                                                                             |
+| FLEET-OBS-009 | not-started | Open PR → green CI (queue guard) → squash-merge → close `fleet-observability-1`                                    | FLEET-OBS-008         | lead        | —   | trunk merge; no direct push to main                                                                                           |
+
+## Proposed MVP rollup row (for the MVP orchestrator — not written by this workstream)
+
+```
+| W-FLEET | in-progress | Fleet (agent-session execution layer) | Phase 2/5 | docs/fleet/TASKS.md | observability dogfooded on live stub fleet; control plane rides federation (W1) |
+```

docs/fleet/north-star.md

@@ -0,0 +1,133 @@
+# Mosaic Fleet — North Star
+
+> **Workstream:** W-FLEET (Fleet) under mission `mvp-20260312`
+> **Umbrella:** [docs/MISSION-MANIFEST.md](../MISSION-MANIFEST.md) · [docs/PRD.md](../PRD.md) (Mosaic Stack v0.1.0)
+> **Status:** doctrine — authored 2026-06-20. Owner of this file: Fleet workstream lead.
+> This document does **not** modify the MVP rollup; a rollup row is proposed, not written here.
+
+## Vision
+
+A **customizable, multi-tenant fleet of always-on AI agents** — each defined by role,
+materialized as a durable, joinable runtime session, coordinated by the proven
+orchestrator/worker model, and observable end-to-end across hosts. Coding today;
+finance, analytics, research as roster entries tomorrow — same primitives, different
+roster. The fleet is the **agent-session execution layer** of the Mosaic Stack MVP:
+the thing federation makes reachable across hosts and the webUI/TUI/CLI make visible.
+
+The USC tmux PoC (durable sessions + `agent-send` comms) proved the model. This
+workstream makes it an official, observable, multi-tenant Mosaic Stack capability.
+
+## The Fleet as means of production (bootstrapping)
+
+The Fleet has a **dual role**, and that is the point:
+
+- **As product** — a multi-tenant agent-fleet capability of Mosaic Stack (this workstream).
+- **As means of production** — the orchestrator/worker fleet that _actually builds the
+  entire MVP_ (federation W1, webUI, TUI, CLI, and the Fleet itself).
+
+We are **building the system that builds the system.** Every other MVP workstream is
+delivered _by_ the fleet, so fleet observability and control are not merely product
+features — they are the **operational floor of the whole delivery effort**. If we cannot
+see and steer the agents, we cannot trust what they ship. This is why Phase 2
+(observability) leads: it is the instrument panel for the factory, dogfooded on the live
+fleet that is, recursively, building Mosaic Stack.
+
+The discipline that makes great power safe is the same gate chain the fleet enforces:
+independent review before merge, green CI, honest completion, decide-and-inform cadence,
+and no irreversible action without authority. The bootstrap is only as trustworthy as
+those gates.
+
+## Alignment with MVP cross-cutting requirements
+
+The Fleet inherits — does not re-invent — the MVP's hard requirements:
+
+| MVP req                       | What it means for the Fleet                                                                                             |
+| ----------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
+| MVP-X1 three-surface parity   | fleet observability/control reachable via **CLI + TUI + webUI** (CLI first; webUI is required for parity, not optional) |
+| MVP-X2 multi-tenant isolation | one tenant = one **Linux uid** (own `systemd --user`, socket, `~/.config/mosaic`); no cross-tenant leakage              |
+| MVP-X3 auth (BetterAuth/SSO)  | operator→fleet and cross-host views are auth-gated through the platform's existing auth                                 |
+| MVP-X4 quality gates          | `pnpm typecheck`/`lint`/`format:check` green before any push                                                            |
+| MVP-X5 federated topology     | cross-host fleet visibility rides the **federation** boundary (W1), not a bespoke broker                                |
+| MVP-X6 OTEL tracing           | heartbeats, sends, and lifecycle events emit spans; `traceparent` crosses the federation boundary                       |
+| MVP-X7 trunk merge            | branch from `main`, squash-merge via PR, never push to `main`                                                           |
+
+## The stack — where every concern lives
+
+One **definition** is the source of truth; the **session** is how it runs.
+
+| Layer                            | Owner                                                                                       | Phase-2 reality                                        | Destination                                             |
+| -------------------------------- | ------------------------------------------------------------------------------------------- | ------------------------------------------------------ | ------------------------------------------------------- |
+| **Definition + identity + auth** | gateway / `mosaic-as` (scoped tokens, #541)                                                 | `roster.yaml` (tenant-tagged)                          | one definition; `mosaic agent --new` materializes it    |
+| **Tenancy boundary**             | **Linux uid per tenant** (linger, own `systemd --user`, own socket, own `~/.config/mosaic`) | one tenant: `jarvis` = tenant zero                     | uid-per-tenant; federation aggregates across hosts      |
+| **Runtime**                      | per-tenant tmux session on isolated socket                                                  | dogfood stub sessions (live now on `mosaic-factory`)   | claude/codex/pi/opencode TUIs                           |
+| **Liveness**                     | **heartbeat protocol** every runtime answers                                                | protocol defined + dogfood stub answers it             | all runtimes answer; "healthy" ≠ "pane alive"           |
+| **Observation**                  | read-only `watch` (native tmux) + `pipe-pane` stream                                        | CLI `watch`/`ps`; explicit opt-in `attach` for control | + auth-gated webUI streams                              |
+| **Control plane**                | **federation** across hosts × tenants                                                       | records already carry `tenant_id` + `host`             | federated gateways expose fleet state; webUI in Phase 5 |
+
+## Operating model (inherited, not reinvented)
+
+The AI-guide law stands: one accountable **orchestrator**, isolated **workers** that
+stop at PR-open, the serialized **gate chain** (independent review → green CI →
+diff-sanity → squash-merge → verify), **decide-and-inform** cadence, and a durable
+**board** so missions survive session death. The Fleet is the infrastructure _under_
+this model. See `mosaicstack-aiguide` whitepapers 01 (inter-agent comms) and 03
+(orchestration model) for the rationale.
+
+## Invariants — "maximal vision, incremental delivery, zero foreclosure"
+
+Every artifact, starting Phase 2, MUST:
+
+1. Carry **`tenant_id` + `host`** in schema and message addressing — even with one of each today.
+2. Treat **isolation socket ≠ invisibility** — anything isolated is surfaced by one command.
+3. Define **healthy = answered a heartbeat within N seconds**, never just "pane alive".
+4. Make **observation read-only by default**; control is an explicit, separate, opt-in verb.
+
+## Observation model
+
+| Verb                                | Behavior                                                                                           |
+| ----------------------------------- | -------------------------------------------------------------------------------------------------- |
+| `mosaic fleet ps`                   | one table joining systemd + tmux + process + idle + last-heartbeat, with drift + boot-enable flags |
+| `mosaic agent watch <name>`         | **read-only** join (grouped session / `-r`), no resize tyranny, no keystrokes                      |
+| `mosaic agent attach <name>`        | explicit interactive takeover (the only path that can type)                                        |
+| `mosaic agent send <name> --verify` | confirms message **accepted**, not merely keystroke-injected                                       |
+
+> Why the current PoC blocks observation: sessions live on the isolated `mosaic-factory`
+> socket (invisible to default `tmux ls`), the only sanctioned read is `capture-pane`
+> (blank for full-screen TUIs), and `attach` is read-write + resizes the session. The
+> verbs above restore "join and observe" safely.
+
+## Phased roadmap
+
+| Phase                  | Outcome                                                                                                                                      | Status  |
+| ---------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- | ------- |
+| 0–1                    | tmux PoC, hardening, published CLI v0.0.34 (#565–#568)                                                                                       | ✅ done |
+| **2 — Observability**  | `fleet ps` (host+tenant aware join), heartbeat protocol + dogfood stub answers it, `agent watch` (read-only), `agent send --verify` receipts | ▶ now   |
+| 3 — Real runtimes      | claude/codex/pi/opencode answer heartbeat; **hybrid lifecycle** (core always-on: orchestrator+reviewer; ephemeral workers per lane)          | planned |
+| 4 — Unified definition | one agent schema in gateway; `mosaic agent --new` → materialized per-tenant session; uid-tenant provisioning                                 | planned |
+| 5 — Control plane      | federation-backed cross-host × cross-tenant fleet view; **webUI** (surface chosen then) for MVP-X1 parity                                    | planned |
+
+## Decisions of record (2026-06-20, with Jason)
+
+- Agent model: **config defines, session runs** (gateway = definition/identity/auth; tmux = runtime).
+- Tenancy: **multi-tenant from the start**; isolation = **per-tenant Linux uid**.
+- Health: **heartbeat required** (dogfood stub implements the protocol now).
+- Lifecycle: **hybrid** — core always-on + ephemeral workers per lane.
+- Observation: **read-only default, opt-in takeover**.
+- Multi-host: **designed-for from day one**; control plane **rides federation (W1)**.
+- Delivery: **CLI-first now**, dogfood against the live stub fleet; webUI deferred to Phase 5.
+- Runtimes: fleet agents default to **Codex / pi-on-Codex**; **Claude is reserved for Claude
+  Code only** (avoid alternate-harness API pricing). Validated durable recipe:
+  `mosaic yolo pi --model openai-codex/gpt-5.5:high`. Durable detached launch requires the
+  runtime-bin on PATH (baked into the pane command) + boot-survival (`enable` + linger),
+  which `fleet init` should automate.
+
+## Assumptions (veto-able)
+
+- `ASSUMPTION:` first-class runtimes = claude, codex, pi, opencode; a "role" (analyst,
+  finance, researcher) = persona + skills + tools on top of a runtime, shipped as a
+  starter role library in the framework.
+- `ASSUMPTION:` the cross-host control plane is the **federation** layer (W1), not a
+  separate `fleetd` daemon.
+- `ASSUMPTION:` Fleet is workstream **W-FLEET** under `mvp-20260312`; a rollup row in
+  `docs/TASKS.md` and a workstream declaration in `MISSION-MANIFEST.md` are proposed to
+  the MVP orchestrator, not written by this workstream.

docs/scratchpads/fleet-observability-phase2.md

@@ -0,0 +1,100 @@
+# Scratchpad — Fleet Phase 2: Observability (W-FLEET)
+
+> Append-only. Mission `mvp-20260312` / workstream W-FLEET.
+> Lead: Jarvis (Claude) at `W-jarvis:mos-claude-18`. Coordinating with `jwoltje@dragon-lin:coder0-0`.
+
+## Mission prompt (2026-06-20)
+
+Establish the north star for the Mosaic Fleet feature and prepare Phase-2 observability
+for delivery. The USC tmux PoC is the proven base. Jason granted lead authority:
+"The fleet is a great way to actually build the MVP — we are building the system that
+builds the system." Dogfood actual agent construction + ad-hoc deployment; coordinate
+with a second agent on `dragon-lin`.
+
+## Decisions of record (with Jason, 2026-06-20)
+
+- Agent model: config defines, session runs (gateway = definition/identity/auth; tmux = runtime).
+- Tenancy: multi-tenant from the start; isolation = per-tenant Linux uid.
+- Health: heartbeat required; dogfood stub implements protocol now.
+- Lifecycle: hybrid (core always-on + ephemeral workers).
+- Observation: read-only default, opt-in takeover.
+- Multi-host: designed-for day one; control plane rides federation (W1), not a bespoke broker.
+- Delivery: CLI-first, dogfood on the live stub fleet; webUI deferred to Phase 5.
+- Fleet is dual-role: product AND means of production (bootstrapping the MVP).
+- Code review = **dual-engine**: Claude **and** gpt-5.5/Codex, run together (Jason: the
+  combination produces the best results). Launch reviewers via `mosaic yolo pi` / `codex`
+  (proven path) or `~/.config/mosaic/tools/codex/codex-code-review.sh`. Applies to all
+  code-review gates incl. FLEET-OBS-008. Per Jason 2026-06-20.
+- Worktree discipline: do fleet work in `~/src/mosaicstack-stack-worktrees/<branch>`, NOT
+  the shared main checkout — concurrent processes mutate `main` there (learned 2026-06-20).
+
+## Environment facts (verified 2026-06-20)
+
+- Fleet is live on `W-jarvis` (uid 1000, `jarvis`, `Linger=yes`) on tmux socket
+  `mosaic-factory`: `_holder`, `canary-pi`, `dogfood-coder`, `dogfood-orchestrator`,
+  `dogfood-reviewer`. All panes run `~/.config/mosaic/fleet/dogfood-agent.py` (stub),
+  including `canary-pi` (roster says runtime=pi → **drift**).
+- Holder + `mosaic-agent@*` units are `active (exited)` but `UnitFileState=disabled`
+  (reboot loses fleet → boot-enable gap to surface).
+- Observation blocked by: isolated socket (hidden from default `tmux ls`), `capture-pane`
+  blank for TUIs, `attach` being read-write + resizing.
+- Second agent: `jwoltje@dragon-lin`, session `coder0-0` (group `coder0`), running `node`,
+  default socket. ssh forward reach confirmed.
+
+## Governance / collision-safety
+
+- `mosaicstack-stack` has active mission `mvp-20260312` with single-writer locks on
+  `docs/MISSION-MANIFEST.md`, `docs/TASKS.md`, `docs/scratchpads/mvp-20260312.md`.
+- This workstream touches NONE of those. All Fleet docs scoped under `docs/fleet/` +
+  this scratchpad. Rollup row proposed, not written.
+
+## Session log
+
+- 2026-06-20: Researched AI guide + fleet code + live state. Established north star with
+  Jason (8 forks decided). Branched `feat/fleet-observability`. Persisted
+  `docs/fleet/{north-star.md,PRD.md,TASKS.md}` + this scratchpad. Next: establish comms
+  with dragon-lin coder, commit docs, begin Phase-2 delivery (heartbeat + `fleet ps`).
+- 2026-06-20 (session 2): Built Phase-2 CLI via worker (commit ab47831): `fleet ps`,
+  `agent watch`, `agent send --verify`, 62 tests. LIVE-verified `fleet ps` on
+  mosaic-factory — correctly flagged canary-pi DRIFT + BOOT-ENABLE, tenant_id+host in JSON.
+  Heartbeat responder added to dogfood-agent.py (FLEET-OBS-002) — `fleet ps` HB now
+  `healthy` for all 4 agents.
+- Coordination: dual-engine-reviewed (Claude+Codex) and merged framework PRs #572
+  (sanitization gate) + #575 (CONSTITUTION extraction) as Lead. Codex caught an Alpine
+  blocker on #572 (refuted by CI); Claude caught a CI-breaking format failure on #575.
+- **FINDINGS (north-star / Phase-3 blockers):**
+  1. Ad-hoc `mosaic yolo {codex,pi}` via `start-agent-session.sh` DIE immediately in a
+     detached tmux pane (codex: "stdin is not a terminal"; pi: same). Only the python stub
+     survives. => Real runtimes have NEVER run durably in the fleet. Launch path (PATH/TTY
+     in the detached shell) must be fixed before Phase-3 real-runtime swap. `fleet ps`
+     caught both dead panes instantly (tool validated).
+  2. `MOSAIC_AGENT_NAME` (set in systemd EnvironmentFile) is NOT propagated into tmux's
+     global env, so agents defaulted to `unknown`. Worked around in dogfood-agent.py via
+     tmux session-name fallback; the systemd/tmux env handoff needs a real fix.
+- Next: rebase on merged main, open Phase-2 PR, dual-engine review, merge, close
+  `fleet-observability-1`. Defer launch-path + env-propagation fixes to Phase 3.
+- 2026-06-21 (session 3): Phase-2 PR #579 merged (3 dual-engine rounds hardened
+  verify+watch). Then closed the launch-path question with Jason's input — CORRECTING
+  earlier findings:
+  - The ad-hoc launch deaths were NOT a fundamental TTY blocker: (a) codex was a stale
+    version (Jason updated it); (b) pi was misconfigured to Claude auth (Jason removed it;
+    default is now Codex). The REAL durable-launch bug is **PATH**: the detached tmux
+    launch shell is login+non-interactive, so it misses `~/.npm-global/bin` (added only in
+    `~/.bashrc`) -> `mosaic: command not found` (127) -> pane dies. tmux panes inherit the
+    tmux _server_ env, so PATH must be baked into the pane command.
+  - **Durable real-agent recipe (validated live on gpt-5.5, Claude-free):**
+    `mosaic yolo pi --model openai-codex/gpt-5.5:high` — pi tolerates detached tmux; a raw
+    interactive TUI (codex CLI) exits without an attached client. Status line confirmed
+    `(openai-codex) gpt-5.5 • high`.
+  - PATH fix landed in `start-agent-session.sh` (commit 32efc13, branch
+    feat/fleet-launch-path): derive runtime-bin prefix (MOSAIC_RUNTIME_BIN | npm prefix |
+    ~/.npm-global/bin | ~/.local/bin), bake `export PATH=...; exec <cmd>` into the pane;
+    `exec` also fixes the drift false-positive. Live-tested under stripped PATH -> durable.
+  - Boot-survival: Jason ran `systemctl --user enable` (+ linger). TODO: auto-enable in
+    **fleet init** so operators never have to remember it (agentic-enhancement cycle).
+  - Future custom Pi harness build: pi cannot self-report its model (track
+    runtime/model/effort as fleet metadata); drift detection should recognize `node` as
+    pi's pane command (a node-wrapped pane can currently read as drift).
+  - Findings recorded in AI Guide playbooks/tmux-fleet.md (aiguide PR #7, merged).
+  - Policy: avoid Claude outside Claude Code (API pricing for alt-harness use) — fleet
+    runtimes default to Codex / pi-on-Codex; Claude stays in Claude Code only.

packages/mosaic/framework/defaults/AGENTS.md

@@ -70,6 +70,9 @@ Skills, hooks, MCP, and plugins are force multipliers you MUST use when applicab
 ## Missing core file
 
 If `CONSTITUTION.md`, `AGENTS.md`, `SOUL.md`, or the runtime contract is missing, stop and report it.
+This agent-facing strictness is intentional and stricter than the launcher: the launcher injects
+`CONSTITUTION.md` tolerantly (skipping it if absent so pre-upgrade hosts keep working), but once a host
+is re-seeded a genuinely missing core file is a stop-and-report condition — not something to proceed past.
 
 ## Session Closure
 

packages/mosaic/framework/defaults/CONSTITUTION.md

@@ -2,8 +2,11 @@
 
 The irreducible, non-negotiable law for every Mosaic agent on every harness.
 
-**Framework-owned.** This file is overwritten verbatim on every upgrade — do not edit it. To change
-behavior, add a `.local.md` overlay or a `policy/` file (tighten-only; see `constitution/LAYER-MODEL.md`).
+**Framework-owned.** This file is overwritten verbatim on every upgrade — do not edit it. There is
+**no `CONSTITUTION.local.md`**: hard gates are not locally overridable. A lower layer may only make
+behavior _stricter_, never relax or override a gate (see Precedence). Operator customization lives in
+other layers — `SOUL.md` / `USER.md` and the tighten-only overlays `STANDARDS.local.md` /
+`SOUL.local.md` / `USER.local.md` / `policy/*.md` (see `constitution/LAYER-MODEL.md`).
 Authored in **capability verbs**: where a gate names a capability ("structured reasoning", "queue
 guard"), the runtime adapter binds it to a concrete tool and states whether absence is a hard stop.
 

packages/mosaic/framework/fleet/examples/coding.yaml

@@ -0,0 +1,32 @@
+version: 1
+transport: tmux
+tmux:
+  socket_name: mosaic-factory
+  holder_session: _holder
+defaults:
+  working_directory: ~
+runtimes:
+  claude:
+    reset_command: /clear
+  pi:
+    reset_command: /new
+agents:
+  - name: orchestrator
+    runtime: claude
+    class: orchestrator
+    persistent_persona: true
+  - name: coder0
+    runtime: pi
+    class: implementer
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true
+  - name: coder1
+    runtime: pi
+    class: implementer
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true
+  - name: reviewer
+    runtime: pi
+    class: reviewer
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true

packages/mosaic/framework/fleet/examples/general.yaml

@@ -0,0 +1,22 @@
+version: 1
+transport: tmux
+tmux:
+  socket_name: mosaic-factory
+  holder_session: _holder
+defaults:
+  working_directory: ~
+runtimes:
+  claude:
+    reset_command: /clear
+  pi:
+    reset_command: /new
+agents:
+  - name: orchestrator
+    runtime: claude
+    class: orchestrator
+    persistent_persona: true
+  - name: generalist
+    runtime: pi
+    class: worker
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true

packages/mosaic/framework/fleet/examples/hybrid.yaml

@@ -0,0 +1,32 @@
+version: 1
+transport: tmux
+tmux:
+  socket_name: mosaic-factory
+  holder_session: _holder
+defaults:
+  working_directory: ~
+runtimes:
+  claude:
+    reset_command: /clear
+  pi:
+    reset_command: /new
+agents:
+  - name: orchestrator
+    runtime: claude
+    class: orchestrator
+    persistent_persona: true
+  - name: coder0
+    runtime: pi
+    class: implementer
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true
+  - name: researcher0
+    runtime: pi
+    class: researcher
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true
+  - name: reviewer
+    runtime: pi
+    class: reviewer
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true

packages/mosaic/framework/fleet/examples/research.yaml

@@ -0,0 +1,32 @@
+version: 1
+transport: tmux
+tmux:
+  socket_name: mosaic-factory
+  holder_session: _holder
+defaults:
+  working_directory: ~
+runtimes:
+  claude:
+    reset_command: /clear
+  pi:
+    reset_command: /new
+agents:
+  - name: orchestrator
+    runtime: claude
+    class: orchestrator
+    persistent_persona: true
+  - name: researcher0
+    runtime: pi
+    class: researcher
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true
+  - name: researcher1
+    runtime: pi
+    class: researcher
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true
+  - name: analyst
+    runtime: pi
+    class: analyst
+    model_hint: openai-codex/gpt-5.5:high
+    reset_between_tasks: true

packages/mosaic/framework/tools/fleet/start-agent-session.sh

@@ -6,6 +6,8 @@ MOSAIC_TMUX_SOCKET=${MOSAIC_TMUX_SOCKET:-mosaic-factory}
 MOSAIC_AGENT_RUNTIME=${MOSAIC_AGENT_RUNTIME:-pi}
 MOSAIC_AGENT_WORKDIR=${MOSAIC_AGENT_WORKDIR:-$HOME}
 MOSAIC_AGENT_COMMAND=${MOSAIC_AGENT_COMMAND:-}
+MOSAIC_HEARTBEAT_RUN_DIR=${MOSAIC_HEARTBEAT_RUN_DIR:-$HOME/.config/mosaic/fleet/run}
+MOSAIC_HEARTBEAT_INTERVAL=${MOSAIC_HEARTBEAT_INTERVAL:-15}
 
 if [ -z "$AGENT_NAME" ]; then
   echo "ERROR: agent name argument or MOSAIC_AGENT_NAME is required" >&2
@@ -26,5 +28,125 @@ if [ -z "$MOSAIC_AGENT_COMMAND" ]; then
   MOSAIC_AGENT_COMMAND="mosaic yolo $MOSAIC_AGENT_RUNTIME"
 fi
 
+# ── Derive a runtime-bin PATH prefix ─────────────────────────────────────────
+# Precedence:
+#   1. $MOSAIC_RUNTIME_BIN (explicit override)
+#   2. $(npm config get prefix)/bin  (if npm is on PATH)
+#   3. Fallbacks: $HOME/.npm-global/bin and $HOME/.local/bin
+#
+# Only directories that already exist are included.  The prefix is baked into
+# the pane command regardless of what the LAUNCHER process's $PATH contains,
+# because the tmux pane inherits the tmux SERVER environment (not this script's
+# environment).  A dir on the launcher's PATH may be absent from the server PATH,
+# so every existing candidate must always be included.  Dedup within the
+# constructed prefix avoids listing the same dir twice.
+_build_runtime_bin_prefix() {
+  local candidates=()
+
+  if [ -n "${MOSAIC_RUNTIME_BIN:-}" ]; then
+    candidates+=("$MOSAIC_RUNTIME_BIN")
+  fi
+
+  if command -v npm >/dev/null 2>&1; then
+    local npm_prefix
+    npm_prefix=$(npm config get prefix 2>/dev/null) || true
+    if [ -n "$npm_prefix" ]; then
+      candidates+=("${npm_prefix}/bin")
+    fi
+  fi
+
+  candidates+=("$HOME/.npm-global/bin")
+  candidates+=("$HOME/.local/bin")
+
+  local prefix=""
+  for dir in "${candidates[@]}"; do
+    [ -d "$dir" ] || continue
+    if [ -z "$prefix" ]; then
+      prefix="$dir"
+    else
+      case ":${prefix}:" in
+        *":${dir}:"*) ;;  # already in our prefix — skip
+        *) prefix="${prefix}:${dir}" ;;
+      esac
+    fi
+  done
+
+  printf '%s' "$prefix"
+}
+
+MOSAIC_RUNTIME_BIN_PREFIX=$(_build_runtime_bin_prefix)
+
+# ── Build the pane command ────────────────────────────────────────────────────
+# The pane command must:
+#   - Export the augmented PATH so the runtime binary is found.
+#   - exec the agent command so the runtime is the pane's foreground process
+#     (makes `fleet ps` pane_current_command check reliable; no DRIFT false-positive).
+#
+# Quoting strategy: single-quote the inner shell snippet so that variable
+# references in MOSAIC_AGENT_COMMAND are NOT expanded here — they expand inside
+# the pane shell.  However, MOSAIC_RUNTIME_BIN_PREFIX and PATH must be expanded
+# NOW (in this script) because the pane shell inherits the tmux server
+# environment, not this script's env.
+#
+# We build the snippet as a double-quoted here-string embedded in a printf call
+# to avoid nested quoting problems.
+
+if [ -n "$MOSAIC_RUNTIME_BIN_PREFIX" ]; then
+  PANE_SHELL_SNIPPET="export PATH=\"${MOSAIC_RUNTIME_BIN_PREFIX}:\${PATH}\"; exec ${MOSAIC_AGENT_COMMAND}"
+else
+  PANE_SHELL_SNIPPET="exec ${MOSAIC_AGENT_COMMAND}"
+fi
+
 mkdir -p "$MOSAIC_AGENT_WORKDIR"
-exec tmux -L "$MOSAIC_TMUX_SOCKET" new-session -d -s "$AGENT_NAME" -c "$MOSAIC_AGENT_WORKDIR" "$MOSAIC_AGENT_COMMAND"
+
+# ── Launch the tmux session (no exec — we continue to wire the heartbeat) ────
+tmux -L "$MOSAIC_TMUX_SOCKET" new-session -d -s "$AGENT_NAME" -c "$MOSAIC_AGENT_WORKDIR" \
+  bash -c "$PANE_SHELL_SNIPPET"
+
+# ── Resolve the pane PID (retry briefly to let the session initialise) ────────
+PANE_PID=""
+for _retry in 1 2 3 4 5; do
+  PANE_PID=$(tmux -L "$MOSAIC_TMUX_SOCKET" list-panes \
+    -t "=${AGENT_NAME}:0.0" -F '#{pane_pid}' 2>/dev/null || true)
+  [ -n "$PANE_PID" ] && break
+  sleep 0.2
+done
+
+# ── Spawn the heartbeat sidecar (detached, best-effort) ──────────────────────
+# The sidecar writes ~/.config/mosaic/fleet/run/<AGENT>.hb atomically while the
+# pane process is alive, then exits so the file goes stale (fleet ps shows stale
+# then PANE=dead).  It is runtime-agnostic: it only cares about the pane PID.
+_start_heartbeat_sidecar() {
+  local agent="$1"
+  local pane_pid="$2"
+  local run_dir="$3"
+  local interval="$4"
+  local hb_file="${run_dir}/${agent}.hb"
+
+  mkdir -p "$run_dir"
+
+  # Write the sidecar as a self-contained bash one-liner so it carries no
+  # references to any variables from this script's environment.
+  local sidecar_script
+  sidecar_script=$(printf \
+    'hb=%s; pid=%s; iv=%s; mkdir -p "$(dirname "$hb")"; while kill -0 "$pid" 2>/dev/null; do tmp="$hb.tmp.$$"; printf "ts=%%s\npid=%%s\nstatus=ok\n" "$(date +%%Y-%%m-%%dT%%H:%%M:%%S%%z)" "$pid" > "$tmp" && mv "$tmp" "$hb"; sleep "$iv"; done' \
+    "$hb_file" "$pane_pid" "$interval")
+
+  # setsid + disown ensures the sidecar survives this script exiting.
+  # stderr/stdout go to /dev/null; failures are non-fatal.
+  if command -v setsid >/dev/null 2>&1; then
+    setsid bash -c "$sidecar_script" </dev/null >/dev/null 2>&1 &
+  else
+    bash -c "$sidecar_script" </dev/null >/dev/null 2>&1 &
+  fi
+  disown $! 2>/dev/null || true
+}
+
+if [ -n "$PANE_PID" ]; then
+  # Guard: do not let sidecar startup failures abort the launcher (set -e).
+  _start_heartbeat_sidecar "$AGENT_NAME" "$PANE_PID" \
+    "$MOSAIC_HEARTBEAT_RUN_DIR" "$MOSAIC_HEARTBEAT_INTERVAL" || \
+    echo "WARNING: heartbeat sidecar could not be started for $AGENT_NAME" >&2
+else
+  echo "WARNING: could not resolve pane PID for $AGENT_NAME — heartbeat sidecar not started" >&2
+fi

packages/mosaic/framework/tools/fleet/test-start-agent-session.sh

@@ -6,13 +6,26 @@ START="$SCRIPT_DIR/start-agent-session.sh"
 SOCKET="mosaic-agent-test-$RANDOM-$$"
 AGENT="agent-$RANDOM"
 WORKDIR=$(mktemp -d)
-trap 'tmux -L "$SOCKET" kill-server >/dev/null 2>&1 || true; rm -rf "$WORKDIR"' EXIT
+
+# Keep a single cleanup trap that accumulates resources.
+CLEANUP_DIRS=("$WORKDIR")
+CLEANUP_SOCKETS=("$SOCKET")
+trap '_cleanup' EXIT
+_cleanup() {
+  for s in "${CLEANUP_SOCKETS[@]:-}"; do
+    tmux -L "$s" kill-server >/dev/null 2>&1 || true
+  done
+  for d in "${CLEANUP_DIRS[@]:-}"; do
+    rm -rf "$d"
+  done
+}
 
 fail() {
   echo "FAIL: $*" >&2
   exit 1
 }
 
+# ── Test 1: basic session creation with workdir check ─────────────────────────
 MOSAIC_TMUX_SOCKET="$SOCKET" \
 MOSAIC_AGENT_WORKDIR="$WORKDIR" \
 MOSAIC_AGENT_COMMAND='bash --noprofile --norc -i' \
@@ -22,6 +35,7 @@ tmux -L "$SOCKET" has-session -t "=$AGENT:0.0" || fail "agent session was not cr
 actual_dir=$(tmux -L "$SOCKET" display-message -p -t "=$AGENT:0.0" '#{pane_current_path}')
 [ "$actual_dir" = "$WORKDIR" ] || fail "agent workdir mismatch: $actual_dir"
 
+# ── Test 2: idempotency (duplicate start prints 'already running') ─────────────
 MOSAIC_TMUX_SOCKET="$SOCKET" \
 MOSAIC_AGENT_WORKDIR="$WORKDIR" \
 MOSAIC_AGENT_COMMAND='bash --noprofile --norc -i' \
@@ -29,4 +43,310 @@ MOSAIC_AGENT_COMMAND='bash --noprofile --norc -i' \
 
 grep -qF 'already running' /tmp/mosaic-start-agent-idempotent.out || fail "duplicate start was not idempotent"
 
+# ── Test 3: runtime-bin PATH prefix is baked into the pane command ────────────
+#
+# We capture the command the script would hand to tmux by injecting a fake
+# 'tmux' shim into PATH.  The shim:
+#   - Intercepts 'new-session' calls and records its arguments to a file.
+#   - For 'has-session' calls, exits 1 (session does not exist) so the script
+#     proceeds to launch instead of printing "already running".
+#   - For 'list-panes' calls, returns empty so PANE_PID stays unset and the
+#     heartbeat sidecar is NOT spawned (heartbeat is not the focus of this test;
+#     test 6 and 7 cover that path).  This prevents any real-filesystem side
+#     effects or leaked background processes.
+#   - For all other subcommands, exits 0.
+#
+# Assertions:
+#   a) 'export PATH=' with the synthetic MOSAIC_RUNTIME_BIN prefix appears.
+#   b) 'exec' appears so the runtime replaces the wrapper shell.
+#   c) MOSAIC_AGENT_COMMAND with flags is forwarded intact.
+
+FAKE_BIN=$(mktemp -d)
+FAKE_RUNTIME_BIN=$(mktemp -d)
+TMUX_ARGS_FILE=$(mktemp)
+HB_RUN_DIR3=$(mktemp -d)
+CLEANUP_DIRS+=("$FAKE_BIN" "$FAKE_RUNTIME_BIN" "$HB_RUN_DIR3")
+
+# Write the fake tmux shim (uses only positional args, no sourced vars).
+cat > "$FAKE_BIN/tmux" <<SHIM
+#!/usr/bin/env bash
+# Fake tmux: record new-session args; report has-session as missing.
+subcmd="\$3"  # argv: tmux -L <socket> <subcmd> ...
+if [ "\$subcmd" = "has-session" ]; then
+  exit 1   # session not found → script will attempt new-session
+fi
+if [ "\$subcmd" = "new-session" ]; then
+  printf '%s\n' "\$@" > "$TMUX_ARGS_FILE"
+  exit 0
+fi
+if [ "\$subcmd" = "list-panes" ]; then
+  # Return empty: no sidecar spawned (heartbeat is not the focus of this test).
+  echo ""
+  exit 0
+fi
+exit 0
+SHIM
+chmod +x "$FAKE_BIN/tmux"
+
+SOCKET3="mosaic-agent-test3-$RANDOM-$$"
+AGENT3="agent3-$RANDOM"
+WORKDIR3=$(mktemp -d)
+CLEANUP_DIRS+=("$WORKDIR3")
+
+PATH="$FAKE_BIN:$PATH" \
+MOSAIC_TMUX_SOCKET="$SOCKET3" \
+MOSAIC_AGENT_WORKDIR="$WORKDIR3" \
+MOSAIC_AGENT_RUNTIME="pi" \
+MOSAIC_RUNTIME_BIN="$FAKE_RUNTIME_BIN" \
+MOSAIC_AGENT_COMMAND="mosaic yolo pi --model openai-codex/gpt-5.5:high" \
+MOSAIC_HEARTBEAT_RUN_DIR="$HB_RUN_DIR3" \
+  "$START" "$AGENT3"
+
+all_args=$(cat "$TMUX_ARGS_FILE" 2>/dev/null || true)
+rm -f "$TMUX_ARGS_FILE"
+
+echo "--- captured tmux new-session args ---"
+echo "$all_args"
+echo "--- end args ---"
+
+# a) PATH prefix containing FAKE_RUNTIME_BIN must appear.
+echo "$all_args" | grep -qF "export PATH=" || fail "pane command does not export PATH"
+echo "$all_args" | grep -qF "$FAKE_RUNTIME_BIN" || fail "pane command does not include MOSAIC_RUNTIME_BIN in PATH prefix"
+
+# b) exec must appear so the runtime replaces the wrapper shell.
+echo "$all_args" | grep -qF "exec " || fail "pane command does not use exec"
+
+# c) Full MOSAIC_AGENT_COMMAND (with flags) must be forwarded.
+echo "$all_args" | grep -qF "mosaic yolo pi --model openai-codex/gpt-5.5:high" || \
+  fail "pane command does not forward MOSAIC_AGENT_COMMAND with flags intact"
+
+# ── Test 4: when no extra runtime-bin dirs exist, exec still appears ───────────
+TMUX_ARGS_FILE2=$(mktemp)
+FAKE_BIN2=$(mktemp -d)
+HB_RUN_DIR4=$(mktemp -d)
+CLEANUP_DIRS+=("$FAKE_BIN2" "$HB_RUN_DIR4")
+
+cat > "$FAKE_BIN2/tmux" <<SHIM2
+#!/usr/bin/env bash
+subcmd="\$3"
+if [ "\$subcmd" = "has-session" ]; then exit 1; fi
+if [ "\$subcmd" = "new-session" ]; then
+  printf '%s\n' "\$@" > "$TMUX_ARGS_FILE2"
+  exit 0
+fi
+if [ "\$subcmd" = "list-panes" ]; then
+  # Return empty: no sidecar spawned (heartbeat is not the focus of this test).
+  echo ""
+  exit 0
+fi
+exit 0
+SHIM2
+chmod +x "$FAKE_BIN2/tmux"
+
+SOCKET4="mosaic-agent-test4-$RANDOM-$$"
+AGENT4="agent4-$RANDOM"
+WORKDIR4=$(mktemp -d)
+CLEANUP_DIRS+=("$WORKDIR4")
+
+# MOSAIC_RUNTIME_BIN points to a non-existent dir so prefix will be empty;
+# .npm-global/bin and .local/bin may or may not exist but we just want exec.
+PATH="$FAKE_BIN2:$PATH" \
+MOSAIC_TMUX_SOCKET="$SOCKET4" \
+MOSAIC_AGENT_WORKDIR="$WORKDIR4" \
+MOSAIC_AGENT_RUNTIME="pi" \
+MOSAIC_RUNTIME_BIN="/nonexistent-dir-$$" \
+MOSAIC_AGENT_COMMAND="mosaic yolo pi" \
+MOSAIC_HEARTBEAT_RUN_DIR="$HB_RUN_DIR4" \
+  "$START" "$AGENT4"
+
+all_args4=$(cat "$TMUX_ARGS_FILE2" 2>/dev/null || true)
+rm -f "$TMUX_ARGS_FILE2"
+rm -rf "$WORKDIR4"
+
+echo "$all_args4" | grep -qF "exec " || fail "pane command (no prefix dirs) does not use exec"
+echo "$all_args4" | grep -qF "mosaic yolo pi" || fail "pane command does not include agent command when no prefix"
+
+# ── Test 5: candidate dir already in LAUNCHER $PATH is still baked into pane ──
+#
+# Regression guard for the bug where _build_runtime_bin_prefix() used to skip
+# a candidate because it was already present in the launcher process's $PATH.
+# That check was wrong: the pane inherits the tmux SERVER environment, not the
+# launcher's env.  Even if a dir is on the launcher's PATH it must always be
+# baked into the pane's PATH export.
+#
+# We prove this by setting PATH to include FAKE_RUNTIME_BIN5 (the candidate),
+# then asserting the generated new-session command still exports it.
+TMUX_ARGS_FILE5=$(mktemp)
+FAKE_BIN5=$(mktemp -d)
+FAKE_RUNTIME_BIN5=$(mktemp -d)  # this dir IS on the launcher's PATH below
+HB_RUN_DIR5=$(mktemp -d)
+CLEANUP_DIRS+=("$FAKE_BIN5" "$FAKE_RUNTIME_BIN5" "$HB_RUN_DIR5")
+
+cat > "$FAKE_BIN5/tmux" <<SHIM5
+#!/usr/bin/env bash
+subcmd="\$3"
+if [ "\$subcmd" = "has-session" ]; then exit 1; fi
+if [ "\$subcmd" = "new-session" ]; then
+  printf '%s\n' "\$@" > "$TMUX_ARGS_FILE5"
+  exit 0
+fi
+if [ "\$subcmd" = "list-panes" ]; then
+  # Return empty: no sidecar spawned (heartbeat is not the focus of this test).
+  echo ""
+  exit 0
+fi
+exit 0
+SHIM5
+chmod +x "$FAKE_BIN5/tmux"
+
+SOCKET5="mosaic-agent-test5-$RANDOM-$$"
+AGENT5="agent5-$RANDOM"
+WORKDIR5=$(mktemp -d)
+CLEANUP_DIRS+=("$WORKDIR5")
+CLEANUP_SOCKETS+=("$SOCKET5")
+
+# FAKE_RUNTIME_BIN5 is deliberately placed on the LAUNCHER PATH so that the
+# old (buggy) code would have skipped it.  The correct code must still include
+# it in the pane PATH export.
+PATH="$FAKE_BIN5:$FAKE_RUNTIME_BIN5:$PATH" \
+MOSAIC_TMUX_SOCKET="$SOCKET5" \
+MOSAIC_AGENT_WORKDIR="$WORKDIR5" \
+MOSAIC_AGENT_RUNTIME="pi" \
+MOSAIC_RUNTIME_BIN="$FAKE_RUNTIME_BIN5" \
+MOSAIC_AGENT_COMMAND="mosaic yolo pi" \
+MOSAIC_HEARTBEAT_RUN_DIR="$HB_RUN_DIR5" \
+  "$START" "$AGENT5"
+
+all_args5=$(cat "$TMUX_ARGS_FILE5" 2>/dev/null || true)
+rm -f "$TMUX_ARGS_FILE5"
+rm -rf "$WORKDIR5"
+
+echo "--- test 5: launcher-PATH candidate must still appear in pane export ---"
+echo "$all_args5"
+echo "--- end test 5 args ---"
+
+echo "$all_args5" | grep -qF "export PATH=" || \
+  fail "test5: pane command does not export PATH when candidate is on launcher PATH"
+echo "$all_args5" | grep -qF "$FAKE_RUNTIME_BIN5" || \
+  fail "test5: candidate dir (already on launcher PATH) was NOT baked into pane PATH — regression"
+
+# ── Test 6: heartbeat sidecar — pane PID resolved + .hb file written ──────────
+#
+# Uses a real tmux session (same socket as test 1 which already has $AGENT) so
+# list-panes returns a real pane PID.  We override MOSAIC_HEARTBEAT_RUN_DIR to
+# a temp dir and set a 1-second interval, then wait up to 3 s for the .hb file
+# to appear and check its content.
+
+HB_RUN_DIR=$(mktemp -d)
+CLEANUP_DIRS+=("$HB_RUN_DIR")
+
+# Re-use the session+agent created in Test 1 (still alive on $SOCKET / $AGENT).
+# We need to invoke the script for a NEW agent on the same socket to exercise
+# the heartbeat path with a real pane PID.
+AGENT6="agent6-$RANDOM"
+MOSAIC_TMUX_SOCKET="$SOCKET" \
+MOSAIC_AGENT_WORKDIR="$WORKDIR" \
+MOSAIC_AGENT_COMMAND='bash --noprofile --norc -i' \
+MOSAIC_HEARTBEAT_RUN_DIR="$HB_RUN_DIR" \
+MOSAIC_HEARTBEAT_INTERVAL="1" \
+  "$START" "$AGENT6"
+
+HB_FILE="$HB_RUN_DIR/${AGENT6}.hb"
+
+# Wait up to 5 seconds for the heartbeat file to appear.
+_waited=0
+until [ -f "$HB_FILE" ] || [ "$_waited" -ge 5 ]; do
+  sleep 0.5
+  _waited=$((_waited + 1))
+done
+
+[ -f "$HB_FILE" ] || fail "test6: heartbeat file not written at $HB_FILE within 5s"
+
+hb_content=$(cat "$HB_FILE")
+echo "--- test 6: heartbeat file content ---"
+echo "$hb_content"
+echo "--- end test 6 ---"
+
+# Verify required fields are present.
+echo "$hb_content" | grep -qE '^ts=[0-9]{4}-[0-9]{2}-[0-9]{2}T' || \
+  fail "test6: heartbeat ts field missing or malformed"
+echo "$hb_content" | grep -qE '^pid=[0-9]+' || \
+  fail "test6: heartbeat pid field missing or malformed"
+echo "$hb_content" | grep -qF 'status=ok' || \
+  fail "test6: heartbeat status=ok missing"
+
+# ── Test 7: heartbeat sidecar — targets correct .hb path per agent name ────────
+#
+# Uses the fake-tmux shim approach (like tests 3-5) to capture the sidecar
+# invocation without needing a real session.  A fake setsid shim records its
+# arguments so we can assert the sidecar script targets the expected .hb path
+# and uses the configured interval.
+
+FAKE_BIN7=$(mktemp -d)
+FAKE_RUNTIME_BIN7=$(mktemp -d)
+SETSID_ARGS_FILE=$(mktemp)
+HB_RUN_DIR7=$(mktemp -d)
+CLEANUP_DIRS+=("$FAKE_BIN7" "$FAKE_RUNTIME_BIN7" "$HB_RUN_DIR7")
+
+AGENT7="my-fleet-agent-$RANDOM"
+INTERVAL7="42"
+
+# Fake tmux: has-session → not found; new-session → ok; list-panes → known PID.
+cat > "$FAKE_BIN7/tmux" <<SHIM7
+#!/usr/bin/env bash
+subcmd="\$3"
+if [ "\$subcmd" = "has-session" ]; then exit 1; fi
+if [ "\$subcmd" = "new-session" ]; then exit 0; fi
+if [ "\$subcmd" = "list-panes" ]; then echo "88888"; exit 0; fi
+exit 0
+SHIM7
+chmod +x "$FAKE_BIN7/tmux"
+
+# Fake setsid: capture the bash -c <script> argument for inspection, then
+# background an actual bash subshell so disown succeeds in the caller.
+cat > "$FAKE_BIN7/setsid" <<'SETSID_SHIM'
+#!/usr/bin/env bash
+# argv: setsid bash -c <sidecar_script>
+# Record the full argument list to the capture file, then exit cleanly.
+printf '%s\0' "$@" > __SETSID_ARGS_FILE__
+exit 0
+SETSID_SHIM
+# Patch the placeholder with the real capture-file path (avoids heredoc expansion issues).
+sed -i "s|__SETSID_ARGS_FILE__|${SETSID_ARGS_FILE}|g" "$FAKE_BIN7/setsid"
+chmod +x "$FAKE_BIN7/setsid"
+
+SOCKET7="mosaic-agent-test7-$RANDOM-$$"
+WORKDIR7=$(mktemp -d)
+CLEANUP_DIRS+=("$WORKDIR7")
+
+PATH="$FAKE_BIN7:$PATH" \
+MOSAIC_TMUX_SOCKET="$SOCKET7" \
+MOSAIC_AGENT_WORKDIR="$WORKDIR7" \
+MOSAIC_AGENT_RUNTIME="pi" \
+MOSAIC_RUNTIME_BIN="$FAKE_RUNTIME_BIN7" \
+MOSAIC_AGENT_COMMAND="mosaic yolo pi" \
+MOSAIC_HEARTBEAT_RUN_DIR="$HB_RUN_DIR7" \
+MOSAIC_HEARTBEAT_INTERVAL="$INTERVAL7" \
+  "$START" "$AGENT7"
+
+# Give the background setsid shim a moment to finish writing the capture file.
+sleep 0.5
+
+setsid_args=$(cat "$SETSID_ARGS_FILE" 2>/dev/null | tr '\0' '\n' || true)
+rm -f "$SETSID_ARGS_FILE"
+rm -rf "$WORKDIR7"
+
+echo "--- test 7: captured setsid args ---"
+echo "$setsid_args"
+echo "--- end test 7 ---"
+
+# The sidecar script (bash -c <script>) must reference the correct .hb path.
+expected_hb="${HB_RUN_DIR7}/${AGENT7}.hb"
+echo "$setsid_args" | grep -qF "$expected_hb" || \
+  fail "test7: sidecar script does not reference correct .hb path ($expected_hb)"
+
+# The sidecar script must use the configured interval.
+echo "$setsid_args" | grep -qF "$INTERVAL7" || \
+  fail "test7: sidecar script does not reference configured interval ($INTERVAL7)"
+
 echo "ok - start-agent-session"

packages/mosaic/framework/tools/quality/scripts/verify-sanitized.sh

@@ -12,7 +12,7 @@
 #   2. STRUCTURAL (private $HOME default in *.sh) — scanned everywhere EXCEPT examples/,
 #        because worked example overlays/personas legitimately show placeholder paths.
 #
-# File types: *.md, *.sh, *.ps1, *.json, and the extensionless CLI scripts under
+# File types: *.md, *.sh, *.ps1, *.json, *.yml/*.yaml, *.toml, *.env, *.service, and the CLI scripts under
 # tools/_scripts/. Excludes node_modules/ and this gate file.
 #
 # NOTE: '\bPDA\b' intentionally matches "PDA-friendly" (the contamination removed in P2);
@@ -39,7 +39,7 @@ cd "$FRAMEWORK_ROOT" || { echo "FRAMEWORK_ROOT not found: $FRAMEWORK_ROOT" >&2;
 # Identity scope = ALL shipped text files (examples/ INCLUDED).
 _files_identity() {
   find . -type f \
-    \( -name '*.md' -o -name '*.sh' -o -name '*.ps1' -o -name '*.json' -o -path '*/tools/_scripts/*' \) \
+    \( -name '*.md' -o -name '*.sh' -o -name '*.ps1' -o -name '*.json' -o -name '*.yml' -o -name '*.yaml' -o -name '*.toml' -o -name '*.env' -o -name '*.service' -o -path '*/tools/_scripts/*' \) \
     -not -path '*/node_modules/*' -not -path "./$SELF_REL" -print0
 }
 # Structural scope = shipped scripts, examples/ EXCLUDED.

packages/mosaic/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mosaicstack/mosaic",
-  "version": "0.0.34",
+  "version": "0.0.36",
   "repository": {
     "type": "git",
     "url": "https://git.mosaicstack.dev/mosaicstack/stack.git",