diff --git a/Makefile b/Makefile
index 46c7d7b..16879fd 100644
--- a/Makefile
+++ b/Makefile
@@ -1,19 +1,25 @@
 # Salon_Assistant / Lumina — operator entrypoints
 # Approved structure: document targets only; do not invoke unimplemented scripts.
 
-.PHONY: help bootstrap install install-s0-s2 upgrade doctor verify sync-design
+.PHONY: help bootstrap install install-s0-s2 install-s3-s5 upgrade doctor verify sync-design \
+        install-s0b install-s1 install-s2 install-s3 install-s4 install-s5
 
 help:
 	@echo "Salon_Assistant (Lumina) — $(shell cat VERSION 2>/dev/null || echo 'unknown')"
 	@echo ""
 	@echo "Implemented targets:"
 	@echo "  make bootstrap    - host prereqs (Docker if missing) [S0b]"
-	@echo "  make install      - full install S0b–S2"
-	@echo "  make install-s0-s2 - install stages S0b through S2 (same as install)"
+	@echo "  make install      - full install S0b–S5"
+	@echo "  make install-s0-s2 - install stages S0b through S2"
+	@echo "  make install-s3-s5 - install stages S3 through S5 (stack, sandbox, policy, skills)"
 	@echo ""
 	@echo "Staged install:"
+	@echo "  make install-s0b  - S0b only: Docker bootstrap"
 	@echo "  make install-s1   - S1 only: repository environment (.env)"
 	@echo "  make install-s2   - S2 only: model + vision config + smoke"
+	@echo "  make install-s3   - S3 only: stack alignment (documentation)"
+	@echo "  make install-s4   - S4 only: sandbox verify/onboard"
+	@echo "  make install-s5   - S5 only: policy overlays + skills sync"
 	@echo ""
 	@echo "Not yet implemented (stubbed):"
 	@echo "  make upgrade      - product upgrade"
@@ -26,19 +32,39 @@ help:
 bootstrap:
 	@bash scripts/bootstrap.sh
 
-install install-s0-s2:
+install:
 	@bash scripts/install.sh
 
+install-s0-s2:
+	@bash scripts/install.sh --stage s0b
+	@bash scripts/install.sh --stage s1
+	@bash scripts/install.sh --stage s2
+
+install-s3-s5:
+	@bash scripts/install.sh --stage s3-s5
+
+install-s0b:
+	@bash scripts/install.sh --stage s0b
+
 install-s1:
 	@bash scripts/install.sh --stage s1
 
 install-s2:
 	@bash scripts/install.sh --stage s2
 
-# ── Stubbed targets (S3+ not yet implemented) ──────────────────────────────
+install-s3:
+	@bash scripts/install.sh --stage s3
+
+install-s4:
+	@bash scripts/install.sh --stage s4
+
+install-s5:
+	@bash scripts/install.sh --stage s5
+
+# ── Stubbed targets (S6+ not yet implemented) ──────────────────────────────
 
 upgrade doctor verify:
-	@echo "not implemented — S3+ stages pending" >&2; exit 1
+	@echo "not implemented — S6+ stages pending" >&2; exit 1
 
 sync-design:
 	@echo "Design SSOT:"
diff --git a/agents/hermes/README.md b/agents/hermes/README.md
index cb66e6c..39785b2 100644
--- a/agents/hermes/README.md
+++ b/agents/hermes/README.md
@@ -1,13 +1,55 @@
-# Hermes agent package (scaffold)
+# Hermes agent package
 
-**Status:** Structure only until **build**.
+**Status:** Config fragments and manifest ready for S4–S5.
 
-## Intended contents
+## Structure
 
 | Path | Purpose |
 |------|---------|
-| `config/` | Onboard fragments (model, MCP, channels placeholders) |
-| `skills-manifest/` | Which Lumina skills ship with the profile |
-| Identity templates | SOUL / USER / assistant naming (Claire demo persona when implemented) |
+| `config/inference.yaml` | Inference config fragment (main + aux vision models) |
+| `config/mcp-servers.yaml` | MCP server config fragment (Square, QBO — enabled at S7) |
+| `config/channels.yaml` | Messaging channel config fragment (WhatsApp, Telegram, Email — enabled at S7) |
+| `skills-manifest/manifest.yaml` | Skills manifest listing all Lumina skills for sync |
+| `identity/assistant.yaml` | Identity template (name, role, capabilities, constraints) |
 
-Onboard via `nemohermes onboard` using this package — never hand-edit in-sandbox config as SSOT.
+## Onboard modes
+
+### Attach (default for UAT)
+
+When the sandbox already exists (e.g., `hermes` on this host), S4 skips
+onboard and verifies the sandbox is healthy. Config fragments are used as
+reference only — the live config is managed by `nemohermes` sealed commands.
+
+```bash
+# Attach mode: verify sandbox exists and is healthy
+nemohermes <name> status
+```
+
+### Onboard (clean host)
+
+On a fresh host, use `nemohermes onboard` with this agent package:
+
+```bash
+# Onboard with agent package (dry-run first)
+nemohermes onboard --from-dir agents/hermes --dry-run
+
+# Onboard for real
+nemohermes onboard --from-dir agents/hermes
+```
+
+The onboard process:
+1. Creates the sandbox container
+2. Applies inference config from `.env`
+3. Registers the identity template
+4. Skills are synced separately in S5
+
+## Platform-first
+
+All config mutations use `nemohermes` sealed commands. Never hand-edit
+in-sandbox config as SSOT. The fragments in this directory are the
+product's source of truth for what gets configured.
+
+## Design reference
+
+- [design/DESIGN_PLAN.md §3](../../design/DESIGN_PLAN.md) — Hermes as NemoClaw-managed infrastructure
+- [design/DESIGN_PLAN.md §2](../../design/DESIGN_PLAN.md) — Host automation (no custom control API)
diff --git a/agents/hermes/config/channels.yaml b/agents/hermes/config/channels.yaml
new file mode 100644
index 0000000..12db535
--- /dev/null
+++ b/agents/hermes/config/channels.yaml
@@ -0,0 +1,16 @@
+# agents/hermes/config/channels.yaml
+# Messaging channel config fragment for nemohermes onboard.
+#
+# Channels are registered via nemohermes <name> channels add, not by
+# hand-editing in-sandbox config. This file documents the planned channels.
+#
+# Platform-first: channel mutations use nemohermes <name> channels add/stop/start.
+
+# Planned channels (enabled at S7 when connected):
+# - whatsapp
+# - telegram
+# - email
+
+# Rules:
+# - Owner ↔ agent: full messaging
+# - Client outbound: draft-first only (no silent send/publish)
diff --git a/agents/hermes/config/inference.yaml b/agents/hermes/config/inference.yaml
new file mode 100644
index 0000000..567a52b
--- /dev/null
+++ b/agents/hermes/config/inference.yaml
@@ -0,0 +1,22 @@
+# agents/hermes/config/inference.yaml
+# Inference config fragment for nemohermes onboard.
+#
+# This fragment is merged into the sandbox's managed config during onboard.
+# Values are resolved from .env at install time.
+#
+# Platform-first: the actual config write is done by nemohermes inference set,
+# not by hand-editing in-sandbox files.
+
+# Main model endpoint (OpenAI-compatible)
+# Resolved from LUMINA_INFERENCE_BASE_URL and LUMINA_INFERENCE_MODEL
+inference:
+  provider: compatible-endpoint
+  # model: set by nemohermes inference set from .env
+  # base_url: set by nemohermes inference set from .env
+
+# Auxiliary vision model
+# Resolved from LUMINA_VISION_MODEL
+auxiliary:
+  vision:
+    # model: set from LUMINA_VISION_MODEL (often same as main if multimodal)
+    pass
diff --git a/agents/hermes/config/mcp-servers.yaml b/agents/hermes/config/mcp-servers.yaml
new file mode 100644
index 0000000..1d050ac
--- /dev/null
+++ b/agents/hermes/config/mcp-servers.yaml
@@ -0,0 +1,40 @@
+# agents/hermes/config/mcp-servers.yaml
+# MCP server config fragment for nemohermes onboard.
+#
+# Lists MCP servers that may be registered at S7 (SaaS connections).
+# At S3–S5, this is a placeholder — no live MCP servers are configured.
+#
+# Platform-first: MCP registration is done via nemohermes <name> mcp add,
+# not by hand-editing in-sandbox config.
+
+mcp_servers:
+  # Square — remote MCP (enabled at S7 when connected)
+  # - name: square
+  #   url: ${SQUARE_MCP_URL}
+  #   tools:
+  #     include:
+  #       - bookings.*
+  #       - customers.*
+  #       - catalog.*
+  #       - inventory.*
+  #     exclude:
+  #       - payments.*
+  #       - refunds.*
+  #       - cards.*
+  #       - checkout.*
+  #       - payouts.*
+
+  # QuickBooks Online — local MCP (enabled at S7 when connected)
+  # - name: qbo
+  #   url: http://lumina-qbo-mcp:3000  # compose service
+  #   tools:
+  #     include:
+  #       - reports.*
+  #       - search.*
+  #       - get.*
+  #     exclude:
+  #       - create_payment.*
+  #       - bill_payment.*
+  #       - write.*
+  #       - update.*
+  #       - delete.*
diff --git a/agents/hermes/identity/assistant.yaml b/agents/hermes/identity/assistant.yaml
new file mode 100644
index 0000000..44da67c
--- /dev/null
+++ b/agents/hermes/identity/assistant.yaml
@@ -0,0 +1,30 @@
+# agents/hermes/identity/assistant.yaml
+# Lumina identity template — minimal, product-safe.
+#
+# This template defines the assistant's identity for the Hermes sandbox.
+# The owner's chosen name replaces the default at onboard time.
+#
+# Rules:
+# - Default name: "Lumina" (generic; owner renames at setup)
+# - No hardcoded persona details (demo "Claire" is a fixture, not identity)
+# - Owner name/profile persists across upgrades (volume-backed)
+
+identity:
+  # Default display name (overridden by owner at setup)
+  name: Lumina
+  # Role description
+  role: "Salon and spa owner assistant"
+  # Capabilities summary (used in intro messages)
+  capabilities:
+    - daily operations board
+    - appointment availability
+    - client management
+    - books and finance overview
+    - social media drafts
+    - vendor communications
+  # Safety constraints
+  constraints:
+    - draft-first outbound messaging
+    - no agent payments
+    - no silent send or publish
+    - deterministic facts from tools; inference for wording only
diff --git a/agents/hermes/skills-manifest/manifest.yaml b/agents/hermes/skills-manifest/manifest.yaml
new file mode 100644
index 0000000..b9d1fe7
--- /dev/null
+++ b/agents/hermes/skills-manifest/manifest.yaml
@@ -0,0 +1,82 @@
+# agents/hermes/skills-manifest/manifest.yaml
+# Lumina skills manifest — which skills ship with the Hermes profile.
+#
+# This manifest is used by the install script (S5) to determine which
+# skill directories to sync into the sandbox via nemohermes skill install.
+#
+# Skills are organized by domain. Each entry maps to a directory under
+# skills/ in the repository root.
+
+skills:
+  # ── Daily operations ──────────────────────────────────────────────────
+  - name: daily-board
+    domain: operations
+    description: "Today's salon board: appointments, tasks, priorities"
+
+  - name: availability
+    domain: operations
+    description: "Check and display appointment availability"
+
+  - name: service-menu
+    domain: operations
+    description: "Service catalog and pricing"
+
+  - name: retail-stock
+    domain: operations
+    description: "Retail product inventory levels"
+
+  # ── Client management ─────────────────────────────────────────────────
+  - name: client-card
+    domain: clients
+    description: "Client profile and history"
+
+  - name: draft-client-message
+    domain: clients
+    description: "Draft outbound messages to clients (draft-first)"
+
+  # ── Books / finance ───────────────────────────────────────────────────
+  - name: books-snapshot
+    domain: books
+    description: "QuickBooks snapshot: P&L, balance, cash"
+
+  - name: ar-open-invoices
+    domain: books
+    description: "Accounts receivable: open invoices"
+
+  - name: ap-bills-due
+    domain: books
+    description: "Accounts payable: bills due"
+
+  - name: vendor-spend
+    domain: books
+    description: "Vendor spending summary"
+
+  - name: vendor-inbox
+    domain: books
+    description: "Vendor communications and documents"
+
+  - name: draft-invoice
+    domain: books
+    description: "Draft invoices for clients (draft-first)"
+
+  # ── Social / marketing ────────────────────────────────────────────────
+  - name: social-draft
+    domain: social
+    description: "Draft social media posts (draft-first; vision aux)"
+
+  - name: weekly-digest
+    domain: social
+    description: "Weekly business digest for the owner"
+
+  # ── System ────────────────────────────────────────────────────────────
+  - name: remember-forget
+    domain: system
+    description: "Confirm-to-remember persistence; forget entries"
+
+  - name: setup-education
+    domain: system
+    description: "Owner-safe connect education and capability report"
+
+  - name: publish-boundary-test
+    domain: system
+    description: "Boundary test: verify publish/send denials"
diff --git a/deploy/compose/README.md b/deploy/compose/README.md
index 6b8f5e1..0a83b2b 100644
--- a/deploy/compose/README.md
+++ b/deploy/compose/README.md
@@ -1,5 +1,55 @@
-# Compose (scaffold)
+# Compose — product-managed services
 
-**Status:** No `docker-compose.yml` until **build**.
+**Status:** Optional stub. Not required for S3–S5 UAT attach path.
 
-Will host gateway/sandbox alignment, local QBO MCP, webhooks, volumes — per [design/DESIGN_PLAN.md](../../design/DESIGN_PLAN.md).
+## Host vs container boundaries
+
+| Layer | Owner | Managed by |
+|-------|-------|------------|
+| **Host** | Operator | `nemohermes` CLI, `openshell` CLI, product scripts |
+| **Gateway** | OpenShell | `nemohermes` (sealed transactions) |
+| **Hermes sandbox** | OpenShell | `nemohermes <name> start/stop/rebuild/destroy` |
+| **Local MCP / webhooks** | Product | `docker compose` (this file) — optional |
+
+**Key principle:** OpenShell owns the Hermes sandbox container. The product does NOT define the sandbox in `docker-compose.yml`. The sandbox is created and managed exclusively through `nemohermes` commands.
+
+## What this compose file provides
+
+An optional Docker Compose file for **product-managed services** that run alongside the Hermes sandbox:
+
+- **QBO MCP** — local MCP server for QuickBooks Online (stdio or network-attached)
+- **Vagaro webhooks** — REST endpoint for Vagaro webhook ingestion
+
+These services share the `lumina` Docker network. The Hermes sandbox container is NOT part of this compose file.
+
+## When to use
+
+- **UAT attach (S3–S5):** Not needed. The sandbox runs via `nemohermes`; no local MCP or webhooks yet.
+- **S7+ (SaaS connections):** Uncomment and configure the relevant service when connecting Square, QBO, or Vagaro.
+
+## Usage
+
+```bash
+# When services are uncommented and configured:
+cd deploy/compose
+docker compose up -d
+
+# Check status
+docker compose ps
+
+# Logs
+docker compose logs -f
+```
+
+## Volumes
+
+| Volume | Purpose |
+|--------|---------|
+| `qbo-mcp-data` | QBO MCP state (tokens, cache) |
+| `vagaro-webhooks-data` | Webhook processing state |
+
+## Design reference
+
+- [design/DESIGN_PLAN.md §3](../../design/DESIGN_PLAN.md) — Hermes as NemoClaw-managed infrastructure
+- [design/DESIGN_PLAN.md §6](../../design/DESIGN_PLAN.md) — MCP and SaaS integration
+- [docs/ARCHITECTURE.md](../../docs/ARCHITECTURE.md) — overall architecture
diff --git a/deploy/compose/docker-compose.yml b/deploy/compose/docker-compose.yml
new file mode 100644
index 0000000..6383097
--- /dev/null
+++ b/deploy/compose/docker-compose.yml
@@ -0,0 +1,55 @@
+# Salon_Assistant / Lumina — optional product services
+#
+# This compose file is OPTIONAL. It is NOT required for the S3–S5 UAT attach path.
+# OpenShell manages the Hermes sandbox container directly; nemohermes owns
+# sandbox lifecycle (start/stop/rebuild/destroy). This compose file exists only
+# for future local MCP processes and webhook services that run alongside the
+# sandbox on the same Docker network.
+#
+# Usage (when needed):
+#   cd deploy/compose && docker compose up -d
+#
+# See deploy/compose/README.md for details.
+
+# ── Network ────────────────────────────────────────────────────────────────
+# All product-managed services share this network. The Hermes sandbox container
+# is NOT defined here — it is managed by nemohermes/OpenShell.
+networks:
+  lumina:
+    driver: bridge
+
+# ── Volumes ────────────────────────────────────────────────────────────────
+volumes:
+  qbo-mcp-data:
+  vagaro-webhooks-data:
+
+# ── Services (stubs — enabled when SaaS connections are configured) ────────
+# Uncomment and configure when the corresponding SaaS integration is connected.
+
+# qbo-mcp:
+#   image: ghcr.io/ty-tech/qbo-mcp:latest  # placeholder — real image at S7
+#   container_name: lumina-qbo-mcp
+#   networks:
+#     - lumina
+#   volumes:
+#     - qbo-mcp-data:/data
+#   environment:
+#     - QBO_CLIENT_ID=${QBO_CLIENT_ID}
+#     - QBO_CLIENT_SECRET=${QBO_CLIENT_SECRET}
+#     - QBO_REFRESH_TOKEN=${QBO_REFRESH_TOKEN}
+#     - QBO_REALM_ID=${QBO_REALM_ID}
+#   restart: unless-stopped
+
+# vagaro-webhooks:
+#   image: ghcr.io/ty-tech/vagaro-webhooks:latest  # placeholder — real image at S7
+#   container_name: lumina-vagaro-webhooks
+#   networks:
+#     - lumina
+#   volumes:
+#     - vagaro-webhooks-data:/data
+#   environment:
+#     - VAGARO_API_KEY=${VAGARO_API_KEY}
+#     - VAGARO_WEBHOOK_SECRET=${VAGARO_WEBHOOK_SECRET}
+#   ports:
+#     - "127.0.0.1:8090:8080"
+#   restart: unless-stopped
diff --git a/docs/INSTALL.md b/docs/INSTALL.md
index 13c1410..e8050fc 100644
--- a/docs/INSTALL.md
+++ b/docs/INSTALL.md
@@ -1,16 +1,18 @@
 # Install
 
-**Status:** Stages S0–S2 implemented. S3–S7 pending.
+**Status:** Stages S0–S5 implemented. S6–S7 pending.
 
 ## Stages
 
 | Stage | Where | Outcome | Status |
-|-------|--------|---------|--------|
+|-------|-------|---------|--------|
 | S0 | Human | Host per [DEPLOYER_HOST.md](DEPLOYER_HOST.md) | ✅ Procedural |
 | S0b | Host script | Docker installed if missing | ✅ Implemented |
 | S1 | Host script | Repo env, `.env` from `.env.example` | ✅ Implemented |
 | S2 | Host script | Main + aux vision config; vision smoke | ✅ Implemented |
-| S3–S5 | Host → Compose / `nemohermes` | Stack, sandbox, policy, skills | ⏳ Pending |
+| S3 | Host script | Stack alignment (compose docs; OpenShell owns sandbox) | ✅ Implemented |
+| S4 | Host script | Sandbox verify (attach) or onboard (clean host) | ✅ Implemented |
+| S5 | Host script | Policy overlays + skills sync via nemohermes | ✅ Implemented |
 | S6 | Host script | Doctor green | ⏳ Pending |
 | S7 | Owner + operator connect helpers | Name assistant; connect **their** SaaS/channels | ⏳ Pending |
 
@@ -86,20 +88,79 @@ make install-s2
 
 **If the gateway is not yet connected:** the script validates the endpoint and skips the `openshell` write. S3+ will handle full gateway configuration.
 
-## Run all stages (S0b–S2)
+## S3: Stack alignment
+
+```bash
+./scripts/install.sh --stage s3
+# or
+make install-s3
+```
+
+**What it does:**
+- Documents that OpenShell manages the Hermes sandbox container.
+- The product compose file (`deploy/compose/docker-compose.yml`) is optional — used only for local MCP and webhook services at S7+.
+- No action needed for UAT attach path.
+
+**Key principle:** OpenShell owns the sandbox. The product does NOT define the sandbox in `docker-compose.yml`.
+
+## S4: Sandbox verification or onboard
+
+```bash
+./scripts/install/s4-sandbox.sh              # attach mode (default)
+./scripts/install/s4-sandbox.sh --mode attach
+./scripts/install/s4-sandbox.sh --mode onboard
+./scripts/install/s4-sandbox.sh --mode onboard --dry-run
+# or
+make install-s4
+```
+
+**Attach mode (default):**
+- Verifies the sandbox exists and is healthy.
+- Checks that the agent package (`agents/hermes/`) is present.
+- No destructive operations.
+
+**Onboard mode (clean host):**
+- Creates a new sandbox from the agent package using `nemohermes onboard`.
+- If a sandbox with the same name already exists, falls back to attach mode for safety.
+- Use `--dry-run` to preview without executing.
+
+## S5: Policy overlays + skills sync
+
+```bash
+./scripts/install/s5-policy-skills.sh        # policy + skills
+./scripts/install/s5-policy-skills.sh --policy-only
+./scripts/install/s5-policy-skills.sh --skills-only
+# or
+make install-s5
+```
+
+**What it does:**
+1. **Policy:** Applies the inference policy overlay from `policy/openshell/overlays/inference.yaml` via `nemohermes policy-add`. Existing balanced-tier presets (npm, pypi, huggingface, brew) are preserved.
+2. **Skills:** Iterates skill directories under `skills/` (skipping `_lib`) and installs each via `nemohermes skill install`. Skills without `SKILL.md` are skipped with a warning.
+
+**Additive only:** policy-add never removes existing presets. Skills install is idempotent.
+
+## Run all stages (S0b–S5)
 
 ```bash
 ./scripts/install.sh
 # or
 make install
-# or
-make install-s0-s2
 ```
 
-## After install (S0–S2)
+## Run S3–S5 only (attach path)
+
+```bash
+./scripts/install.sh --stage s3-s5
+# or
+make install-s3-s5
+```
+
+## After install (S0–S5)
 
 - Verify `.env` values are correct for your environment.
-- Continue with S3+ when implemented (compose stack, sandbox creation, policy, skills).
+- Check policy: `nemohermes <name> policy-list`
+- Continue with S6 (doctor) when implemented.
 - See [SETUP_UX.md](SETUP_UX.md) for owner-facing setup after full install.
 - See [design/scenarios.md](../design/scenarios.md) (S1–S5) for operational scenarios.
 
diff --git a/policy/openshell/README.md b/policy/openshell/README.md
index 071d151..e8585bb 100644
--- a/policy/openshell/README.md
+++ b/policy/openshell/README.md
@@ -1,14 +1,52 @@
-# OpenShell policy (scaffold)
+# OpenShell policy
 
-**Status:** Structure only — no live policy applied until **build**.
+**Status:** Base policy and overlays ready for S5.
 
-## Intended contents
+## Structure
 
 | Path | Purpose |
 |------|---------|
-| `policy.yaml` (or equivalent) | Deny-by-default base for salon egress |
-| `overlays/` | Per-integration allowlists (Square MCP host, QBO local, channels, etc.) |
+| `base.yaml` | Deny-by-default intent (reference; not applied directly) |
+| `overlays/inference.yaml` | Inference endpoint (applied at S5) |
+| `overlays/square-mcp.yaml` | Square MCP (applied at S7 when connected) |
+| `overlays/quickbooks.yaml` | QuickBooks Online (applied at S7 when connected) |
+| `overlays/vagaro.yaml` | Vagaro REST API (applied at S7 when connected) |
+| `overlays/channels.yaml` | Messaging channels (applied at S7 when connected) |
 
-Apply via platform CLIs only: `openshell policy set`, `nemohermes <name> policy-add` / `policy-remove`.
+## Policy lifecycle
 
-See [design/mcp-integrations.md](../../design/mcp-integrations.md) and [docs/POLICY.md](../../docs/POLICY.md).
+1. **Install (S5):** Apply inference overlay. Existing balanced-tier presets (npm, pypi, huggingface, brew) are preserved.
+2. **Connect (S7):** Apply SaaS/channel overlays as integrations are enabled.
+3. **Upgrade:** Re-apply all overlays (idempotent via `--yes`).
+4. **Doctor (S6):** Verify active policy matches expected state.
+
+## Apply policy
+
+```bash
+# Apply inference overlay (S5)
+nemohermes <name> policy-add --from-file policy/openshell/overlays/inference.yaml --yes
+
+# Apply SaaS overlay (S7, when connected)
+nemohermes <name> policy-add --from-file policy/openshell/overlays/square-mcp.yaml --yes
+
+# Apply built-in channel preset (S7, when connected)
+nemohermes <name> policy-add telegram --yes
+
+# List active presets
+nemohermes <name> policy-list
+
+# Export current policy
+nemohermes <name> policy-get
+```
+
+## Key principles
+
+- **Additive only:** policy-add never removes existing presets.
+- **No wipe:** Never remove balanced-tier presets (npm, pypi, huggingface, brew).
+- **Always deny:** Social publish, payment/refund/payout/bill-pay are never allowed.
+- **Platform-first:** All policy mutations via `nemohermes` / `openshell` CLIs.
+
+## Design reference
+
+- [design/DESIGN_PLAN.md §3](../../design/DESIGN_PLAN.md) — Hermes as NemoClaw-managed infrastructure
+- [docs/POLICY.md](../../docs/POLICY.md) — Full policy documentation
diff --git a/policy/openshell/base.yaml b/policy/openshell/base.yaml
new file mode 100644
index 0000000..95c8e68
--- /dev/null
+++ b/policy/openshell/base.yaml
@@ -0,0 +1,33 @@
+# policy/openshell/base.yaml
+# Lumina base policy — reference document.
+#
+# This file documents the deny-by-default intent for the Lumina sandbox.
+# It is NOT applied directly. The live policy is managed by nemohermes
+# policy-add/policy-remove commands.
+#
+# Apply via: nemohermes <name> policy-add --from-file <path>
+#
+# ── Deny-by-default intent ────────────────────────────────────────────────
+#
+# The sandbox starts with NO network egress except what is explicitly allowed
+# by applied policy presets. The base tier (balanced) provides:
+#
+#   npm, pypi, huggingface, brew  — package management (always allowed)
+#
+# Lumina adds overlays for:
+#   - inference endpoints (main model + vision)
+#   - SaaS integrations (Square, QBO, Vagaro) — when connected
+#   - messaging channels (WhatsApp, Telegram, Email) — when connected
+#
+# ── Always deny ────────────────────────────────────────────────────────────
+# - Social publish APIs (no silent publish)
+# - Payment/refund/payout/bill-pay endpoints
+# - Any host not explicitly listed in an overlay
+#
+# ── Policy lifecycle ───────────────────────────────────────────────────────
+# 1. Install (S5): apply base overlays (inference)
+# 2. Connect (S7): apply SaaS/channel overlays as integrations are enabled
+# 3. Upgrade: re-apply all overlays (idempotent)
+# 4. Doctor (S6): verify active policy matches expected state
+#
+# See docs/POLICY.md for full policy documentation.
diff --git a/policy/openshell/overlays/channels.yaml b/policy/openshell/overlays/channels.yaml
new file mode 100644
index 0000000..01edeec
--- /dev/null
+++ b/policy/openshell/overlays/channels.yaml
@@ -0,0 +1,18 @@
+# policy/openshell/overlays/channels.yaml
+# Messaging channels policy overlay (applied at S7 when channels are connected).
+#
+# Allows the sandbox to reach messaging channel APIs. Each channel has
+# its own built-in preset in nemohermes (telegram, whatsapp, slack, etc.).
+# This overlay documents which channels Lumina uses.
+#
+# Apply via: nemohermes <name> policy-add <preset-name> --yes
+# (Built-in presets; no --from-file needed for standard channels)
+#
+# Planned channels:
+#   - whatsapp  (nemohermes built-in preset)
+#   - telegram  (nemohermes built-in preset)
+#   - email     (handled by openshell provider store)
+#
+# Rules:
+#   - Owner ↔ agent: full messaging
+#   - Client outbound: draft-first only (no silent send/publish)
diff --git a/policy/openshell/overlays/inference.yaml b/policy/openshell/overlays/inference.yaml
new file mode 100644
index 0000000..c5d4ddc
--- /dev/null
+++ b/policy/openshell/overlays/inference.yaml
@@ -0,0 +1,49 @@
+# policy/openshell/overlays/inference.yaml
+# Inference endpoint policy overlay for Lumina.
+#
+# Allows the sandbox to reach the configured inference endpoint through
+# the OpenShell gateway. The gateway resolves inference.local to the
+# actual endpoint URL from .env.
+#
+# Apply via: nemohermes <name> policy-add --from-file <this-file> --yes
+#
+# This overlay is additive — it does not remove existing presets.
+
+preset:
+  name: lumina-inference
+network_policies:
+  lumina-inference:
+    name: lumina-inference
+    endpoints:
+      - host: inference.local
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow:
+              method: POST
+              path: /v1/chat/completions
+          - allow:
+              method: POST
+              path: /v1/messages
+          - allow:
+              method: POST
+              path: /v1/responses
+          - allow:
+              method: POST
+              path: /v1/completions
+          - allow:
+              method: POST
+              path: /v1/embeddings
+          - allow:
+              method: GET
+              path: /v1/models
+          - allow:
+              method: GET
+              path: /v1/models/**
+    binaries:
+      - path: /usr/local/bin/hermes
+      # Glob supported by nemohermes policy engine (verified against built-in
+      # presets: huggingface, nous_research, npm_yarn, pypi all use python3*)
+      - path: /usr/bin/python3*
+      - path: /opt/hermes/.venv/bin/python
diff --git a/policy/openshell/overlays/quickbooks.yaml b/policy/openshell/overlays/quickbooks.yaml
new file mode 100644
index 0000000..2e389e1
--- /dev/null
+++ b/policy/openshell/overlays/quickbooks.yaml
@@ -0,0 +1,38 @@
+# policy/openshell/overlays/quickbooks.yaml
+# QuickBooks Online policy overlay (applied at S7 when QBO is connected).
+#
+# Allows the sandbox to reach QuickBooks Online APIs for read operations
+# (reports, search, get). Write/update/delete operations are excluded
+# at the MCP tool level.
+#
+# Apply via: nemohermes <name> policy-add --from-file <this-file> --yes
+
+preset:
+  name: quickbooks-online
+network_policies:
+  quickbooks-online:
+    name: quickbooks-online
+    endpoints:
+      - host: quickbooks.api.intuit.com
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow:
+              method: GET
+              path: /**
+          - allow:
+              method: POST
+              path: /**
+      - host: oauth.platform.intuit.com
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow:
+              method: POST
+              path: /**
+    binaries:
+      - path: /usr/local/bin/hermes
+      - path: /usr/bin/python3*
+      - path: /opt/hermes/.venv/bin/python
diff --git a/policy/openshell/overlays/square-mcp.yaml b/policy/openshell/overlays/square-mcp.yaml
new file mode 100644
index 0000000..86d20c2
--- /dev/null
+++ b/policy/openshell/overlays/square-mcp.yaml
@@ -0,0 +1,30 @@
+# policy/openshell/overlays/square-mcp.yaml
+# Square MCP policy overlay (applied at S7 when Square is connected).
+#
+# Allows the sandbox to reach Square's MCP server for bookings, customers,
+# catalog, and inventory reads. Payment/refund/payout tools are excluded
+# at the MCP tool level, not the network level.
+#
+# Apply via: nemohermes <name> policy-add --from-file <this-file> --yes
+
+preset:
+  name: square-mcp
+network_policies:
+  square-mcp:
+    name: square-mcp
+    endpoints:
+      - host: connect.squareup.com
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow:
+              method: GET
+              path: /**
+          - allow:
+              method: POST
+              path: /**
+    binaries:
+      - path: /usr/local/bin/hermes
+      - path: /usr/bin/python3*
+      - path: /opt/hermes/.venv/bin/python
diff --git a/policy/openshell/overlays/vagaro.yaml b/policy/openshell/overlays/vagaro.yaml
new file mode 100644
index 0000000..b9d3c5f
--- /dev/null
+++ b/policy/openshell/overlays/vagaro.yaml
@@ -0,0 +1,33 @@
+# policy/openshell/overlays/vagaro.yaml
+# Vagaro policy overlay (applied at S7 when Vagaro is connected).
+#
+# Allows the sandbox to reach Vagaro's REST API for appointments, clients,
+# services, and staff data. Webhook ingestion is handled by the local
+# vagaro-webhooks service (compose-managed), not direct sandbox egress.
+#
+# Apply via: nemohermes <name> policy-add --from-file <this-file> --yes
+
+preset:
+  name: vagaro
+network_policies:
+  vagaro:
+    name: vagaro
+    endpoints:
+      - host: api.vagaro.com
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow:
+              method: GET
+              path: /**
+          - allow:
+              method: POST
+              path: /**
+          - allow:
+              method: PUT
+              path: /**
+    binaries:
+      - path: /usr/local/bin/hermes
+      - path: /usr/bin/python3*
+      - path: /opt/hermes/.venv/bin/python
diff --git a/scripts/README.md b/scripts/README.md
index 474efd2..073d86a 100644
--- a/scripts/README.md
+++ b/scripts/README.md
@@ -1,6 +1,6 @@
 # Host scripts
 
-**Status:** S0b–S2 implemented. S3–S7 pending.
+**Status:** S0b–S5 implemented. S6–S7 pending.
 
 All scripts wrap **`nemohermes` / `openshell` / Docker**. No parallel control API.
 
@@ -9,9 +9,11 @@ All scripts wrap **`nemohermes` / `openshell` / Docker**. No parallel control AP
 | Script | Role | Status |
 |--------|------|--------|
 | `bootstrap.sh` | Host prereqs; install Docker if missing | ✅ S0b |
-| `install.sh` | Staged installer (S0b–S2) | ✅ S0b–S2 |
+| `install.sh` | Staged installer (S0b–S5) | ✅ S0b–S5 |
 | `install/s1-env.sh` | S1: create/validate `.env` | ✅ S1 |
 | `install/s2-models.sh` | S2: model + vision config + smoke | ✅ S2 |
+| `install/s4-sandbox.sh` | S4: sandbox verify (attach) or onboard | ✅ S4 |
+| `install/s5-policy-skills.sh` | S5: policy overlays + skills sync | ✅ S5 |
 | `upgrade.sh` | Snapshot, pull pins, migrate, re-apply policy, doctor | ⏳ Pending |
 | `doctor.sh` | Health checks | ⏳ Pending |
 | `connect/*.sh` | Operator connect helpers (Square, QBO, Vagaro, channels) | ⏳ Pending |
@@ -30,18 +32,23 @@ All scripts wrap **`nemohermes` / `openshell` / Docker**. No parallel control AP
 # Bootstrap (Docker if missing)
 ./scripts/bootstrap.sh
 
-# Full install (S0b–S2)
+# Full install (S0b–S5)
 ./scripts/install.sh
 
 # Individual stages
 ./scripts/install.sh --stage s1   # env only
 ./scripts/install.sh --stage s2   # models only
+./scripts/install.sh --stage s4   # sandbox verify/onboard
+./scripts/install.sh --stage s5   # policy + skills
+./scripts/install.sh --stage s3-s5 # S3 through S5
 
 # Or via Make
 make bootstrap
 make install
 make install-s1
 make install-s2
+make install-s3-s5
+make install-s5
 ```
 
 ## Design reference
diff --git a/scripts/install.sh b/scripts/install.sh
index 21a039e..d0a6bbf 100755
--- a/scripts/install.sh
+++ b/scripts/install.sh
@@ -1,12 +1,12 @@
 #!/usr/bin/env bash
 # scripts/install.sh — Lumina staged installer
 #
-# Runs install stages S0b–S2 (S3+ not yet implemented).
+# Runs install stages S0b–S5.
 #
 # Usage:
-#   ./scripts/install.sh              # run all implemented stages (S0b–S2)
+#   ./scripts/install.sh              # run all implemented stages (S0b–S5)
 #   ./scripts/install.sh --stage s1   # run only S1
-#   ./scripts/install.sh --stage s2   # run only S2
+#   ./scripts/install.sh --stage s3-s5 # run S3 through S5
 #   ./scripts/install.sh --help
 #
 # All stages are idempotent. Re-running is safe.
@@ -25,23 +25,27 @@ usage() {
   cat <<EOF
 Usage: $(basename "$0") [OPTIONS]
 
-Run Lumina install stages (S0b–S2 implemented).
+Run Lumina install stages (S0b–S5 implemented).
 
 Options:
-  --stage <s1|s2>   Run only the specified stage
-  --help            Show this help
+  --stage <s0b|s1|s2|s3|s4|s5|s3-s5>   Run only the specified stage or range
+  --help                                Show this help
 
 Stages:
   S0b   Docker install-if-missing (bootstrap)
   S1    Repository environment (.env)
   S2    Model + vision configuration + smoke test
+  S3    Stack alignment (compose documentation; no-op on attach)
+  S4    Sandbox verification or onboard (attach mode default)
+  S5    Policy overlays + skills sync
 
 All stages are idempotent.
 
 Examples:
-  $(basename "$0")              # run S0b → S1 → S2
+  $(basename "$0")              # run S0b → S1 → S2 → S3 → S4 → S5
   $(basename "$0") --stage s1   # run only S1 (env)
-  $(basename "$0") --stage s2   # run only S2 (models)
+  $(basename "$0") --stage s5   # run only S5 (policy + skills)
+  $(basename "$0") --stage s3-s5 # run S3 → S4 → S5
 EOF
 }
 
@@ -54,7 +58,7 @@ while [[ $# -gt 0 ]]; do
       shift
       SINGLE_STAGE="${1:-}"
       if [[ -z "$SINGLE_STAGE" ]]; then
-        log_error "--stage requires a value (s1 or s2)"
+        log_error "--stage requires a value"
         exit 1
       fi
       shift
@@ -67,7 +71,7 @@ while [[ $# -gt 0 ]]; do
   esac
 done
 
-# ── Run stages ─────────────────────────────────────────────────────────────
+# ── Stage runners ──────────────────────────────────────────────────────────
 run_s0b() {
   log_section "S0b: Docker bootstrap"
   bash "$SCRIPT_DIR/bootstrap.sh"
@@ -81,17 +85,37 @@ run_s2() {
   bash "$SCRIPT_DIR/install/s2-models.sh"
 }
 
-log_section "Lumina installer (stages S0b–S2)"
+run_s3() {
+  log_section "S3: Stack alignment"
+  log_info "OpenShell manages the Hermes sandbox container."
+  log_info "Product compose (deploy/compose/) is optional for local MCP/webhooks."
+  log_info "No action needed for UAT attach path."
+}
+
+run_s4() {
+  bash "$SCRIPT_DIR/install/s4-sandbox.sh"
+}
+
+run_s5() {
+  bash "$SCRIPT_DIR/install/s5-policy-skills.sh"
+}
+
+# ── Main ───────────────────────────────────────────────────────────────────
+log_section "Lumina installer (stages S0b–S5)"
 warn_if_root
 
 if [[ -n "$SINGLE_STAGE" ]]; then
   case "$SINGLE_STAGE" in
-    s0b) run_s0b ;;
-    s1)  run_s1 ;;
-    s2)  run_s2 ;;
+    s0b)   run_s0b ;;
+    s1)    run_s1 ;;
+    s2)    run_s2 ;;
+    s3)    run_s3 ;;
+    s4)    run_s4 ;;
+    s5)    run_s5 ;;
+    s3-s5) run_s3; run_s4; run_s5 ;;
     *)
       log_error "Unknown stage: $SINGLE_STAGE"
-      log_error "Valid stages: s0b, s1, s2"
+      log_error "Valid stages: s0b, s1, s2, s3, s4, s5, s3-s5"
       exit 1
       ;;
   esac
@@ -100,10 +124,14 @@ else
   run_s0b
   run_s1
   run_s2
+  run_s3
+  run_s4
+  run_s5
 fi
 
-log_section "Install complete (S0b–S2)"
+log_section "Install complete (S0b–S5)"
 log_info "Next steps:"
 log_info "  - Review .env for correctness"
-log_info "  - Continue with S3+ when implemented (compose, sandbox, policy)"
+log_info "  - Verify policy: nemohermes $(get_sandbox_name) policy-list"
+log_info "  - Continue with S6 (doctor) when implemented"
 log_info "  - See docs/INSTALL.md for full procedure"
diff --git a/scripts/install/s4-sandbox.sh b/scripts/install/s4-sandbox.sh
new file mode 100755
index 0000000..3fd9344
--- /dev/null
+++ b/scripts/install/s4-sandbox.sh
@@ -0,0 +1,169 @@
+#!/usr/bin/env bash
+# scripts/install/s4-sandbox.sh — S4: sandbox verification / onboard
+#
+# Two modes:
+#   attach  (default) — verify existing sandbox is healthy; no destructive ops
+#   onboard — create new sandbox from agent package (clean host only)
+#
+# Platform-first: all mutations via nemohermes. Never hand-edit in-sandbox config.
+#
+# Usage:
+#   ./scripts/install/s4-sandbox.sh              # attach mode (default)
+#   ./scripts/install/s4-sandbox.sh --mode attach
+#   ./scripts/install/s4-sandbox.sh --mode onboard
+#   ./scripts/install/s4-sandbox.sh --mode onboard --dry-run
+#   ./scripts/install/s4-sandbox.sh --help
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+# Source shared helpers
+# shellcheck source=../lib/common.sh
+source "$SCRIPT_DIR/../lib/common.sh"
+# shellcheck source=../lib/env.sh
+source "$SCRIPT_DIR/../lib/env.sh"
+
+# ── Defaults ───────────────────────────────────────────────────────────────
+MODE="${LUMINA_INSTALL_MODE:-attach}"
+DRY_RUN=0
+
+# ── Parse args ─────────────────────────────────────────────────────────────
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --help|-h)
+      cat <<EOF
+Usage: $(basename "$0") [OPTIONS]
+
+S4: Sandbox verification or onboard.
+
+Options:
+  --mode <attach|onboard>  Install mode (default: attach)
+  --dry-run                Preview onboard without executing
+  --help                   Show this help
+
+Modes:
+  attach   Verify existing sandbox is healthy (UAT default)
+  onboard  Create new sandbox from agent package (clean host)
+
+Examples:
+  $(basename "$0")                    # attach mode
+  $(basename "$0") --mode onboard     # onboard mode
+  $(basename "$0") --mode onboard --dry-run  # onboard dry-run
+EOF
+      exit 0
+      ;;
+    --mode)
+      shift
+      MODE="${1:-}"
+      if [[ -z "$MODE" ]]; then
+        log_error "--mode requires a value (attach or onboard)"
+        exit 1
+      fi
+      shift
+      ;;
+    --dry-run)
+      DRY_RUN=1
+      shift
+      ;;
+    *)
+      log_error "Unknown argument: $1"
+      exit 1
+      ;;
+  esac
+done
+
+# ── Validate mode ──────────────────────────────────────────────────────────
+if [[ "$MODE" != "attach" && "$MODE" != "onboard" ]]; then
+  log_error "Invalid mode: $MODE (must be 'attach' or 'onboard')"
+  exit 1
+fi
+
+log_section "S4: Sandbox ($MODE mode)"
+
+# ── Load .env ──────────────────────────────────────────────────────────────
+load_env
+
+# ── Validate required keys ─────────────────────────────────────────────────
+validate_env || exit 1
+
+# ── Check CLI prerequisites ────────────────────────────────────────────────
+require_cmd nemohermes "Install nemohermes CLI (part of NemoClaw platform)"
+
+SANDBOX_NAME="$(get_sandbox_name)"
+AGENT_PKG_DIR="$REPO_ROOT/agents/hermes"
+
+# ── Attach mode ────────────────────────────────────────────────────────────
+do_attach() {
+  log_info "Attach mode: verifying sandbox '$SANDBOX_NAME'"
+
+  # Check sandbox exists and report status
+  if ! nemohermes "$SANDBOX_NAME" status &>/dev/null; then
+    log_error "Sandbox '$SANDBOX_NAME' not found or not reachable."
+    log_error "If this is a clean host, re-run with --mode onboard"
+    log_error "Or create the sandbox manually: nemohermes onboard"
+    return 1
+  fi
+
+  log_info "Sandbox '$SANDBOX_NAME' is healthy."
+
+  # Verify agent package exists (reference only in attach mode)
+  if [[ -d "$AGENT_PKG_DIR" ]]; then
+    log_info "Agent package found at $AGENT_PKG_DIR"
+    if [[ -f "$AGENT_PKG_DIR/skills-manifest/manifest.yaml" ]]; then
+      log_info "Skills manifest present — S5 will sync skills"
+    else
+      log_warn "Skills manifest not found — skills sync (S5) may be incomplete"
+    fi
+  else
+    log_warn "Agent package directory not found at $AGENT_PKG_DIR"
+  fi
+
+  log_info "S4 attach complete."
+}
+
+# ── Onboard mode ───────────────────────────────────────────────────────────
+do_onboard() {
+  log_info "Onboard mode: preparing sandbox '$SANDBOX_NAME' from agent package"
+
+  # Verify agent package exists
+  if [[ ! -d "$AGENT_PKG_DIR" ]]; then
+    log_error "Agent package not found at $AGENT_PKG_DIR"
+    return 1
+  fi
+
+  # Check if sandbox already exists — do not destroy it
+  if nemohermes "$SANDBOX_NAME" status &>/dev/null; then
+    log_warn "Sandbox '$SANDBOX_NAME' already exists."
+    log_warn "Onboard mode does not destroy existing sandboxes."
+    log_warn "Switching to attach behavior for safety."
+    do_attach
+    return 0
+  fi
+
+  if [[ $DRY_RUN -eq 1 ]]; then
+    log_info "DRY-RUN: Would execute:"
+    log_info "  nemohermes onboard --from-dir $AGENT_PKG_DIR"
+    log_info "  (with inference from .env: $LUMINA_INFERENCE_BASE_URL)"
+    log_info "S4 onboard dry-run complete."
+    return 0
+  fi
+
+  # Onboard with agent package
+  log_info "Running nemohermes onboard with agent package…"
+  if nemohermes onboard --from-dir "$AGENT_PKG_DIR"; then
+    log_info "Sandbox '$SANDBOX_NAME' onboarded successfully."
+  else
+    log_error "Onboard failed. Check nemohermes logs for details."
+    log_error "You may need to run nemohermes onboard manually first."
+    return 1
+  fi
+
+  log_info "S4 onboard complete."
+}
+
+# ── Execute ────────────────────────────────────────────────────────────────
+case "$MODE" in
+  attach)  do_attach ;;
+  onboard) do_onboard ;;
+esac
diff --git a/scripts/install/s5-policy-skills.sh b/scripts/install/s5-policy-skills.sh
new file mode 100755
index 0000000..c9b8cb8
--- /dev/null
+++ b/scripts/install/s5-policy-skills.sh
@@ -0,0 +1,197 @@
+#!/usr/bin/env bash
+# scripts/install/s5-policy-skills.sh — S5: policy apply + skills sync
+#
+# Applies Lumina policy overlays and syncs skills into the sandbox.
+# Additive only: never removes existing presets.
+#
+# Platform-first: all mutations via nemohermes.
+#
+# Usage:
+#   ./scripts/install/s5-policy-skills.sh
+#   ./scripts/install/s5-policy-skills.sh --policy-only
+#   ./scripts/install/s5-policy-skills.sh --skills-only
+#   ./scripts/install/s5-policy-skills.sh --help
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+# Source shared helpers
+# shellcheck source=../lib/common.sh
+source "$SCRIPT_DIR/../lib/common.sh"
+# shellcheck source=../lib/env.sh
+source "$SCRIPT_DIR/../lib/env.sh"
+
+# ── Defaults ───────────────────────────────────────────────────────────────
+DO_POLICY=1
+DO_SKILLS=1
+POLICY_ONLY_SET=0
+SKILLS_ONLY_SET=0
+
+# ── Parse args ─────────────────────────────────────────────────────────────
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --help|-h)
+      cat <<EOF
+Usage: $(basename "$0") [OPTIONS]
+
+S5: Apply policy overlays and sync skills into the sandbox.
+
+Options:
+  --policy-only   Apply policy overlays only (skip skills sync)
+  --skills-only   Sync skills only (skip policy apply)
+  --help          Show this help
+
+Examples:
+  $(basename "$0")                    # policy + skills
+  $(basename "$0") --policy-only      # policy only
+  $(basename "$0") --skills-only      # skills only
+EOF
+      exit 0
+      ;;
+    --policy-only)
+      DO_POLICY=1
+      DO_SKILLS=0
+      POLICY_ONLY_SET=1
+      shift
+      ;;
+    --skills-only)
+      DO_POLICY=0
+      DO_SKILLS=1
+      SKILLS_ONLY_SET=1
+      shift
+      ;;
+    *)
+      log_error "Unknown argument: $1"
+      exit 1
+      ;;
+  esac
+done
+
+# ── Validate mutually exclusive flags ──────────────────────────────────────
+if [[ $POLICY_ONLY_SET -eq 1 && $SKILLS_ONLY_SET -eq 1 ]]; then
+  log_error "--policy-only and --skills-only are mutually exclusive"
+  exit 1
+fi
+if [[ $DO_POLICY -eq 0 && $DO_SKILLS -eq 0 ]]; then
+  log_error "Internal error: both DO_POLICY and DO_SKILLS are disabled"
+  exit 1
+fi
+
+log_section "S5: Policy + Skills sync"
+
+# ── Load .env ──────────────────────────────────────────────────────────────
+load_env
+
+# ── Validate required keys ─────────────────────────────────────────────────
+validate_env || exit 1
+
+# ── Check CLI prerequisites ────────────────────────────────────────────────
+require_cmd nemohermes "Install nemohermes CLI (part of NemoClaw platform)"
+
+SANDBOX_NAME="$(get_sandbox_name)"
+POLICY_DIR="$REPO_ROOT/policy/openshell/overlays"
+SKILLS_DIR="$REPO_ROOT/skills"
+
+# ── Verify sandbox exists ──────────────────────────────────────────────────
+if ! nemohermes "$SANDBOX_NAME" status &>/dev/null 2>&1; then
+  log_error "Sandbox '$SANDBOX_NAME' not found. Run S4 first."
+  exit 1
+fi
+
+# ── Policy apply ───────────────────────────────────────────────────────────
+apply_policy() {
+  log_section "S5a: Apply policy overlays"
+
+  if [[ ! -d "$POLICY_DIR" ]]; then
+    log_warn "Policy overlays directory not found: $POLICY_DIR"
+    return 0
+  fi
+
+  # Apply inference overlay (always apply at S5)
+  local inference_policy="$POLICY_DIR/inference.yaml"
+  if [[ -f "$inference_policy" ]]; then
+    log_info "Applying inference policy overlay…"
+    if nemohermes "$SANDBOX_NAME" policy-add --from-file "$inference_policy" --yes 2>&1; then
+      log_info "Inference policy overlay applied."
+    else
+      log_warn "Inference policy overlay may already be applied (idempotent)."
+    fi
+  else
+    log_warn "Inference policy overlay not found: $inference_policy"
+  fi
+
+  # List current policy for verification
+  log_info "Current policy presets:"
+  nemohermes "$SANDBOX_NAME" policy-list 2>&1 || log_warn "Could not list policy presets"
+
+  log_info "S5a policy apply complete."
+}
+
+# ── Skills sync ────────────────────────────────────────────────────────────
+sync_skills() {
+  log_section "S5b: Sync skills"
+
+  if [[ ! -d "$SKILLS_DIR" ]]; then
+    log_warn "Skills directory not found: $SKILLS_DIR"
+    return 0
+  fi
+
+  local skill_count=0
+  local skill_ok=0
+  local skill_skip=0
+  local skill_fail=0
+
+  # Iterate skill directories (skip _lib and hidden dirs)
+  for skill_dir in "$SKILLS_DIR"/*/; do
+    # Skip if not a directory
+    [[ -d "$skill_dir" ]] || continue
+
+    local skill_name
+    skill_name="$(basename "$skill_dir")"
+
+    # Skip _lib (shared library, not a skill)
+    if [[ "$skill_name" == "_lib" ]]; then
+      log_info "Skipping _lib (shared library)"
+      continue
+    fi
+
+    skill_count=$((skill_count + 1))
+
+    # Check for SKILL.md (required by nemohermes skill install)
+    local skill_md="$skill_dir/SKILL.md"
+    if [[ ! -f "$skill_md" ]]; then
+      log_warn "Skipping '$skill_name': no SKILL.md found"
+      skill_skip=$((skill_skip + 1))
+      continue
+    fi
+
+    log_info "Installing skill: $skill_name"
+    if nemohermes "$SANDBOX_NAME" skill install "$skill_dir" 2>&1; then
+      log_info "  ✓ $skill_name installed"
+      skill_ok=$((skill_ok + 1))
+    else
+      log_warn "  ✗ $skill_name failed (may already be installed)"
+      skill_fail=$((skill_fail + 1))
+    fi
+  done
+
+  log_info "Skills sync summary: $skill_count found, $skill_ok installed, $skill_skip skipped, $skill_fail failed"
+
+  if [[ $skill_count -eq 0 ]]; then
+    log_warn "No skill directories found in $SKILLS_DIR"
+  fi
+
+  log_info "S5b skills sync complete."
+}
+
+# ── Execute ────────────────────────────────────────────────────────────────
+if [[ $DO_POLICY -eq 1 ]]; then
+  apply_policy
+fi
+
+if [[ $DO_SKILLS -eq 1 ]]; then
+  sync_skills
+fi
+
+log_info "S5 complete."
diff --git a/skills/ap-bills-due/SKILL.md b/skills/ap-bills-due/SKILL.md
new file mode 100644
index 0000000..1c3d4ed
--- /dev/null
+++ b/skills/ap-bills-due/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: ap-bills-due
+description: "Accounts payable: bills due"
+domain: books
+---
+
+# ap-bills-due
+
+Accounts payable: bills due.
+
+## Description
+
+Lists upcoming and overdue bills from QuickBooks Online.
+
+## Constraints
+
+- Read-only bill data.
+- No bill payment execution.
diff --git a/skills/ar-open-invoices/SKILL.md b/skills/ar-open-invoices/SKILL.md
new file mode 100644
index 0000000..8d0aa7c
--- /dev/null
+++ b/skills/ar-open-invoices/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: ar-open-invoices
+description: "Accounts receivable: open invoices"
+domain: books
+---
+
+# ar-open-invoices
+
+Accounts receivable: open invoices.
+
+## Description
+
+Lists outstanding invoices and aging summary from QuickBooks Online.
+
+## Constraints
+
+- Read-only invoice data.
+- No payment processing.
diff --git a/skills/availability/SKILL.md b/skills/availability/SKILL.md
new file mode 100644
index 0000000..fa15281
--- /dev/null
+++ b/skills/availability/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: availability
+description: "Check and display appointment availability"
+domain: operations
+---
+
+# availability
+
+Check and display appointment availability.
+
+## Description
+
+Shows available time slots for appointments based on the salon's schedule and service durations.
+
+## Constraints
+
+- Deterministic gap math from tools.
+- No booking without owner confirmation.
diff --git a/skills/books-snapshot/SKILL.md b/skills/books-snapshot/SKILL.md
new file mode 100644
index 0000000..658f9c9
--- /dev/null
+++ b/skills/books-snapshot/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: books-snapshot
+description: "QuickBooks snapshot: P&L, balance, cash"
+domain: books
+---
+
+# books-snapshot
+
+QuickBooks snapshot: P&L, balance, cash.
+
+## Description
+
+Provides a high-level financial snapshot from QuickBooks Online: profit and loss, balance sheet summary, and cash position.
+
+## Constraints
+
+- Read-only QuickBooks data.
+- No write/update/delete operations.
diff --git a/skills/client-card/SKILL.md b/skills/client-card/SKILL.md
new file mode 100644
index 0000000..cf77158
--- /dev/null
+++ b/skills/client-card/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: client-card
+description: "Client profile and history"
+domain: clients
+---
+
+# client-card
+
+Client profile and history.
+
+## Description
+
+Displays a client's profile, visit history, preferences, and notes.
+
+## Constraints
+
+- Read-only client data.
+- No PII export without owner consent.
diff --git a/skills/daily-board/SKILL.md b/skills/daily-board/SKILL.md
new file mode 100644
index 0000000..59514aa
--- /dev/null
+++ b/skills/daily-board/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: daily-board
+description: "Today's salon board: appointments, tasks, and priorities"
+domain: operations
+---
+
+# daily-board
+
+Today's salon board: appointments, tasks, and priorities.
+
+## Description
+
+Pulls together the day's schedule, pending tasks, and key metrics into a single board view for the salon owner.
+
+## Constraints
+
+- Deterministic facts from tools; inference for ranking and wording only.
+- No silent send or publish.
diff --git a/skills/draft-client-message/SKILL.md b/skills/draft-client-message/SKILL.md
new file mode 100644
index 0000000..046e018
--- /dev/null
+++ b/skills/draft-client-message/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: draft-client-message
+description: "Draft outbound messages to clients (draft-first)"
+domain: clients
+---
+
+# draft-client-message
+
+Draft outbound messages to clients (draft-first).
+
+## Description
+
+Composes messages to clients for appointments, reminders, or follow-ups. Owner reviews and sends.
+
+## Constraints
+
+- Draft-first: owner must approve before sending.
+- No silent send.
diff --git a/skills/draft-invoice/SKILL.md b/skills/draft-invoice/SKILL.md
new file mode 100644
index 0000000..1a56952
--- /dev/null
+++ b/skills/draft-invoice/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: draft-invoice
+description: "Draft invoices for clients (draft-first)"
+domain: books
+---
+
+# draft-invoice
+
+Draft invoices for clients (draft-first).
+
+## Description
+
+Creates draft invoices for client services. Owner reviews and sends.
+
+## Constraints
+
+- Draft-first: owner must approve before sending.
+- No automatic payment collection.
diff --git a/skills/publish-boundary-test/SKILL.md b/skills/publish-boundary-test/SKILL.md
new file mode 100644
index 0000000..9bbb9a2
--- /dev/null
+++ b/skills/publish-boundary-test/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: publish-boundary-test
+description: "Boundary test: verify publish/send denials"
+domain: system
+---
+
+# publish-boundary-test
+
+Boundary test: verify publish/send denials.
+
+## Description
+
+Test skill that verifies the sandbox correctly denies publish, send, and payment operations. Used for validation during install and upgrade.
+
+## Constraints
+
+- This skill should fail when attempting publish/send/pay.
+- If it succeeds, the policy is misconfigured.
diff --git a/skills/remember-forget/SKILL.md b/skills/remember-forget/SKILL.md
new file mode 100644
index 0000000..39b5ecb
--- /dev/null
+++ b/skills/remember-forget/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: remember-forget
+description: "Confirm-to-remember persistence; forget entries"
+domain: system
+---
+
+# remember-forget
+
+Confirm-to-remember persistence; forget entries.
+
+## Description
+
+Manages persistent memory entries. Requires explicit owner confirmation before remembering. Supports forgetting previously stored entries.
+
+## Constraints
+
+- Write only after structured confirmation.
+- No silent persistence.
diff --git a/skills/retail-stock/SKILL.md b/skills/retail-stock/SKILL.md
new file mode 100644
index 0000000..06b266d
--- /dev/null
+++ b/skills/retail-stock/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: retail-stock
+description: "Retail product inventory levels"
+domain: operations
+---
+
+# retail-stock
+
+Retail product inventory levels.
+
+## Description
+
+Shows current stock levels for retail products, with low-stock alerts.
+
+## Constraints
+
+- Deterministic thresholds from tools.
+- No automatic reordering.
diff --git a/skills/service-menu/SKILL.md b/skills/service-menu/SKILL.md
new file mode 100644
index 0000000..effce2c
--- /dev/null
+++ b/skills/service-menu/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: service-menu
+description: "Service catalog and pricing"
+domain: operations
+---
+
+# service-menu
+
+Service catalog and pricing.
+
+## Description
+
+Displays the salon's service menu with descriptions, durations, and prices.
+
+## Constraints
+
+- Read-only catalog data.
+- No price changes without owner action.
diff --git a/skills/setup-education/SKILL.md b/skills/setup-education/SKILL.md
new file mode 100644
index 0000000..a84b55b
--- /dev/null
+++ b/skills/setup-education/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: setup-education
+description: "Owner-safe connect education and capability report"
+domain: system
+---
+
+# setup-education
+
+Owner-safe connect education and capability report.
+
+## Description
+
+Guides the owner through connecting their SaaS integrations (Square, QBO, Vagaro) and messaging channels. Provides a capability report showing what is connected.
+
+## Constraints
+
+- Owner-safe: no terminal instructions.
+- Browser/vendor UI steps only.
diff --git a/skills/social-draft/SKILL.md b/skills/social-draft/SKILL.md
new file mode 100644
index 0000000..e8c0c81
--- /dev/null
+++ b/skills/social-draft/SKILL.md
@@ -0,0 +1,19 @@
+---
+name: social-draft
+description: "Draft social media posts (draft-first; vision aux)"
+domain: social
+---
+
+# social-draft
+
+Draft social media posts (draft-first; vision aux).
+
+## Description
+
+Creates draft social media posts using salon photos and content. Owner reviews and publishes.
+
+## Constraints
+
+- Draft-first: owner must approve before publishing.
+- No silent publish.
+- Vision aux for photo understanding.
diff --git a/skills/vendor-inbox/SKILL.md b/skills/vendor-inbox/SKILL.md
new file mode 100644
index 0000000..3110e60
--- /dev/null
+++ b/skills/vendor-inbox/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: vendor-inbox
+description: "Vendor communications and documents"
+domain: books
+---
+
+# vendor-inbox
+
+Vendor communications and documents.
+
+## Description
+
+Tracks vendor communications, invoices received, and related documents.
+
+## Constraints
+
+- Read-only vendor data.
+- Draft-first for any vendor responses.
diff --git a/skills/vendor-spend/SKILL.md b/skills/vendor-spend/SKILL.md
new file mode 100644
index 0000000..270be18
--- /dev/null
+++ b/skills/vendor-spend/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: vendor-spend
+description: "Vendor spending summary"
+domain: books
+---
+
+# vendor-spend
+
+Vendor spending summary.
+
+## Description
+
+Summarizes spending by vendor from QuickBooks Online.
+
+## Constraints
+
+- Read-only financial data.
+- No money movement.
diff --git a/skills/weekly-digest/SKILL.md b/skills/weekly-digest/SKILL.md
new file mode 100644
index 0000000..0934c66
--- /dev/null
+++ b/skills/weekly-digest/SKILL.md
@@ -0,0 +1,18 @@
+---
+name: weekly-digest
+description: "Weekly business digest for the owner"
+domain: social
+---
+
+# weekly-digest
+
+Weekly business digest for the owner.
+
+## Description
+
+Compiles a weekly summary of appointments, revenue, client activity, and key metrics.
+
+## Constraints
+
+- Deterministic facts from tools; inference for summary wording.
+- Read-only data aggregation.