README.md

@@ -4,7 +4,7 @@ A CLI that turns a **pristine Debian server into a hardened, tailnet-joined
 node** — one curl, one command. A second command installs a version-pinned
 Coolify on a control-plane box.
 
-Philosophy (shared with [claudebox](https://github.com/heavy-duty/claudebox)):
+Philosophy (shared with [box](https://github.com/heavy-duty/box)):
 **public tool, private state**. rig carries plumbing logic only — no
 hostnames, no bindings, no secrets, nothing about *your* infrastructure. It
 takes arguments, does its work, and stores no credential, ever.
@@ -188,6 +188,18 @@ which legitimately lack it. (The world-readable global install path — box unde
 `/opt/box` readable by every non-root user — depends on box PR #71; until that
 merges box's root install lands in `/root`.)
 
+> **The box install is unpinned — on purpose, and out loud.** `coolify install`
+> demands a version pin; the box step tracks a moving `heavy-duty/box@main`.
+> Not because box self-updates (it doesn't — it has Coolify's shape, not the
+> runner's) but because there is nothing to pin *to*: box cuts no tags and no
+> releases, and its installer resolves `refs/heads/<ref>` — branches only — so
+> a `BOX_REF=v0.5.0` would 404 even if the tag existed. Issue #12's call was
+> that silently tracking `main` on the box that runs the agents is the option
+> not to pick — hence this paragraph. `BOX_REPO` / `BOX_REF` are the pin
+> points the day box cuts a tag (or you point at a frozen branch of your own
+> fork); `RIG_SKIP_BOX_INSTALL=1` opts out entirely for a host whose box you
+> manage by hand.
+
 `dev` is `staging`'s human-class sibling — the same VM-hosting, `tag:local`
 shape with a person living on it, box CLI installed the same way — and
 `workstation` is the machine at the keyboard end of all the SSH connections:
@@ -203,6 +215,28 @@ root login is unattributable by construction. `rig users apply` puts named
 operators on every box, server-class included; a human always enters as
 themself and elevates via sudo.
 
+Per role, the whole identity picture at a glance — issue #25's class
+comparison, translated onto the traits that replaced the class binary:
+
+| role            | class  | host | join    | who lives here                       | root SSH after `rig users apply` |
+|-----------------|--------|------|---------|--------------------------------------|----------------------------------|
+| `control-plane` | server | no   | authkey | nobody — Coolify runs here           | open — the automation door       |
+| `workload`      | server | no   | authkey | nobody — deployed services run here  | open — the automation door       |
+| `runner`        | server | no   | authkey | nobody — CI jobs as `github-runner`  | open — the automation door       |
+| `staging`       | server | yes  | authkey | nobody — an unattended VM appliance  | open — the automation door       |
+| `dev`           | human  | yes  | authkey | operators, minting boxes             | closed by `rig users close-root` |
+| `workstation`   | human  | yes  | login   | its owner                            | closed by `rig users close-root` |
+
+Who installs what, and who runs as what: **bootstrap is always root** and
+installs everything a role needs — on `host=yes` that includes the box CLI
+(globally) and box's own `setup-host`. **Humans always run as themselves**:
+operators land via `rig users apply` on every class and elevate through sudo
+(roles `admin`/`rig`) or the `incus` group (role `box`) — never by logging in
+as root. **Machine identities stay machine-shaped**: Coolify's automation
+SSHes in as root (that is what server-class root *is*), CI jobs run as the
+unprivileged `github-runner`, and guest VMs are their own server-class boxes,
+converged from inside by `rig bootstrap workload`.
+
 **`class` decides root SSH's fate — after `rig users apply`, never before.**
 On `class=human`, root SSH closes entirely (`rig users close-root`, below).
 On `class=server` it stays open — key-only, as bootstrap left it — because

commands/bootstrap.sh

@@ -531,7 +531,30 @@ if [ "$HOST" = "yes" ]; then
     # else — a warning, never an abort: box is the host extra, the OS+tailnet core
     # is already done.
     if curl -fsSL "$BOX_INSTALL_URL" | BOX_YES=1 bash; then
-      log "box installed and host set up — mint guest boxes with 'box new'"
+      # Don't trust the exit code — prove the effective state (issue #12). An
+      # installer can exit 0 having done less than it claims: box's setup-host
+      # is written for a sudo-capable user, and one of its paths exits 0 after
+      # only adding a group, asking for a re-login. That is the sshd first-wins
+      # bug's exact shape — asserting what was REQUESTED (here: the installer's
+      # claimed success) instead of what actually TOOK.
+      # Two proofs, one claim each. `command -v box` proves the CLI landed
+      # (box's root install symlinks into /usr/local/bin, already on this
+      # shell's PATH — no login shell needed). "Host set up" is a separate
+      # claim and gets box's OWN effective-state verdict, `box doctor` —
+      # the daemon, the pool, the network stay box's domain (the same
+      # delegation law as the install itself: box owns the daemon), rig
+      # just refuses to claim what that verdict does not answer. Both
+      # failures WARN, never die: box is the host EXTRA, and the OS+tailnet
+      # core above is already done and asserted.
+      if command -v box >/dev/null 2>&1; then
+        if box doctor >/dev/null 2>&1; then
+          log "box installed and host set up — 'box doctor' passed; mint guest boxes with 'box new'"
+        else
+          warn "box is on PATH but 'box doctor' does not pass — the CLI landed, the host stack is unproven. Run 'box doctor' for the verdict, then 'box setup-host' (or finish by hand: ${BOX_MANUAL})"
+        fi
+      else
+        warn "box's installer reported success but no 'box' is on PATH — the install did not take effect. Finish the host by hand: ${BOX_MANUAL}"
+      fi
     else
       warn "box install did not complete (no network, or box's installer failed); bootstrap's core work is done. Finish the host by hand: ${BOX_MANUAL}"
     fi

commands/coolify-backup-install.sh

@@ -7,6 +7,10 @@
 # already-filled bindings file is left untouched.
 set -euo pipefail
 
+HERE="$(cd "$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")" && pwd)"
+# shellcheck source=SCRIPTDIR/lib/users-config.sh
+. "$HERE/lib/users-config.sh"   # read_role_marker — the traits line bootstrap wrote
+
 log()  { printf 'rig-coolify-backup: %s\n' "$*"; }
 warn() { printf 'rig-coolify-backup: WARNING: %s\n' "$*" >&2; }
 die()  { printf 'rig-coolify-backup: ERROR: %s\n' "$1" >&2; exit "${2:-1}"; }
@@ -67,6 +71,18 @@ done
 [ -n "$PG_USER" ]      || die "--pg-user must not be empty" 2
 [ -n "$PG_DB" ]        || die "--pg-db must not be empty" 2
 
+# --- role-marker sanity (issue #25) ------------------------------------------
+# Same advisory check as `rig coolify install`, same reasoning: this command
+# dumps the CONTROL PLANE's database, so a marker naming any other role almost
+# certainly means the wrong SSH session — but the marker is advisory and may be
+# absent, so WARN, never die, and warn before the root check so the harness can
+# prove it non-root (RIG_ROLE_MARKER points it at fixtures, repo precedent).
+MARKER_LINE="$(read_role_marker "${RIG_ROLE_MARKER:-/etc/rig/role}")"
+case "$MARKER_LINE" in
+  ""|"role=control-plane"|"role=control-plane "*) ;;
+  *) warn "this box's role marker says '${MARKER_LINE}' — not a control-plane box. The nightly dump targets Coolify's own database, which lives on role control-plane; if this is the wrong box, stop here and re-check your SSH session." ;;
+esac
+
 # --- guards ----------------------------------------------------------------
 [ "$(id -u)" -eq 0 ] || die "must run as root"
 if [ -r /etc/os-release ]; then

commands/coolify-install.sh

@@ -4,8 +4,13 @@
 # explicit act.
 set -euo pipefail
 
-log() { printf 'rig-coolify: %s\n' "$*"; }
-die() { printf 'rig-coolify: ERROR: %s\n' "$1" >&2; exit "${2:-1}"; }
+HERE="$(cd "$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")" && pwd)"
+# shellcheck source=SCRIPTDIR/lib/users-config.sh
+. "$HERE/lib/users-config.sh"   # read_role_marker — the traits line bootstrap wrote
+
+log()  { printf 'rig-coolify: %s\n' "$*"; }
+warn() { printf 'rig-coolify: WARNING: %s\n' "$*" >&2; }
+die()  { printf 'rig-coolify: ERROR: %s\n' "$1" >&2; exit "${2:-1}"; }
 
 usage() {
   cat <<'EOF'
@@ -32,6 +37,26 @@ if [ -z "$VERSION" ]; then
   die "--version <pin> is required" 2
 fi
 
+# --- role-marker sanity (issue #25) ------------------------------------------
+# Coolify belongs on the control-plane box and nowhere else — but the marker is
+# ADVISORY, never a gate. It may legitimately be absent (a box bootstrapped
+# before rig wrote markers, or a hand-built one), and rig refuses to guess from
+# silence. When the marker EXISTS and names another role, the likeliest story
+# is an operator in the wrong SSH session about to put a control plane on a
+# workload box — so say it loudly. But WARN, never die: the operator may also
+# be deliberately repurposing the box, and an advisory file must never outrank
+# the human running the command (contrast close-root, where the marker IS the
+# gate — shutting the root door blind is irreversible in a way an extra
+# Coolify is not). Placed BEFORE the root check for the same reason arg errors
+# are: the harness proves it non-root, and reading a 0644 file needs no
+# privilege. RIG_ROLE_MARKER overrides the path so tests point it at fixtures
+# (repo precedent: users-apply, users-close-root).
+MARKER_LINE="$(read_role_marker "${RIG_ROLE_MARKER:-/etc/rig/role}")"
+case "$MARKER_LINE" in
+  ""|"role=control-plane"|"role=control-plane "*) ;;
+  *) warn "this box's role marker says '${MARKER_LINE}' — not a control-plane box. Coolify belongs on role control-plane; if this is the wrong box, stop here and re-check your SSH session. Repurposing it on purpose? Re-run 'rig bootstrap control-plane' first so the marker tells the truth." ;;
+esac
+
 [ "$(id -u)" -eq 0 ] || die "must run as root"
 
 export AUTOUPDATE=false

docs/plans/2026-07-18-class-model-remnants.md

@@ -0,0 +1,62 @@
+# Class-model remnants — closing issues #12 and #25
+
+**Goal:** finish the four remnants that keep rig#12 (the `dev` role) and rig#25
+(machine classes) open. The bulk of both issues already landed on `main` via
+PR #27 (traits model: role→class/host/join map, the `dev` and `workstation`
+roles, `/etc/rig/role`, the effective-tag refusals, `rig users`) and PR #28
+(host-class bootstrap installs box + runs its `setup-host`). What remains is
+documentation debt and two small behaviors both issues explicitly asked for.
+
+## Tasks
+
+- [x] **README: `claudebox` → `box`** (#12 comment, item 1). The philosophy
+  line still links `heavy-duty/claudebox`; the repo renamed to
+  `heavy-duty/box`. The redirect works today and is one squatted rename away
+  from not working. Negative-grep test pins the stale slug out.
+- [x] **README: the per-role identity table** (#25, item 4). *The identity
+  model* section carries the prose but not #25's at-a-glance class comparison.
+  Add a compact table translated onto the current traits (class/host/join per
+  role, who lives there, root SSH's fate) plus a who-installs-what /
+  who-runs-as-what paragraph. No wholesale rewrite of the section.
+- [x] **README: say out loud that the box install is unpinned** (#12 comment,
+  item 5). box has no tags/releases and its installer resolves `refs/heads`
+  only, so host-class bootstrap can only track a moving `heavy-duty/box@main`.
+  The issue's decision: install `main` **and the README says so and why** —
+  `BOX_REPO`/`BOX_REF` as the pin points, `RIG_SKIP_BOX_INSTALL=1` as the
+  opt-out.
+- [x] **bootstrap: verify the box install took effect** (#12 comment, item 3 —
+  "rig trusting an exit code instead of checking effective state"). After the
+  installer claims success, prove `command -v box` resolves; a hollow success
+  WARNS (box is the host extra — never fatal) with the manual pointer.
+- [x] **coolify verbs: role-marker sanity warnings** (#25, item 3's named
+  consumer — "`rig <cmd>` sanity warnings later (e.g. `coolify install` on a
+  non-control-plane box)"). Both `coolify install` and `coolify backup
+  install` read `/etc/rig/role` via the lib's `read_role_marker` and warn when
+  it names a non-control-plane role.
+- [x] **Tests** in `test/cli.sh`, existing patterns only: live marker-warning
+  checks through `RIG_ROLE_MARKER` fixtures (non-root), grep-the-shipped-script
+  guards for the root-gated paths, line-number ordering assert for the
+  effective check, fail-closed defaults.
+
+## Behavior contract
+
+- **bootstrap, box block** (`commands/bootstrap.sh`): on the installer-success
+  path, `command -v box` decides the message — found → success log naming
+  `box new` and `box doctor` (box's own effective-state verdict; rig never
+  interrogates Incus, so the deeper verification is delegated, not
+  reimplemented); absent → `warn` with the `BOX_MANUAL` pointer. Never `die`
+  in either branch; exit codes and all skip/failure paths unchanged.
+- **coolify install / coolify backup install**: the marker check is
+  ADVISORY — absent marker or `role=control-plane …` stays silent; any other
+  marker line warns and proceeds. It runs after arg validation and **before**
+  the root check (testable non-root; a 0644 file needs no privilege), reads
+  the path from `RIG_ROLE_MARKER` (default `/etc/rig/role`), and never
+  changes an exit code: usage errors stay 2, the root refusal stays 1.
+- **README**: content-only edits; no command semantics described differently
+  from what ships.
+
+## Non-goals
+
+- No new roles, traits, flags, or marker consumers beyond the two coolify
+  verbs. No changes to box. No gating (warn-only) — `close-root` remains the
+  only command the marker can refuse.

test/cli.sh

@@ -136,6 +136,45 @@ check "bootstrap: box install runs after the role marker write" \
 # host whose box did not install is never left without the next move.
 check "bootstrap: box skip/failure keeps a pointer to the manual install" 0 "" \
   grep -q "prepare Incus" "$ROOT/commands/bootstrap.sh"
+# "Don't trust exit codes" (#12): box's installer can exit 0 having done less
+# than it claims (its setup-host has a path that exits 0 after only adding a
+# group, asking for a re-login). After a claimed success bootstrap must prove
+# the one artifact it asked for — box on PATH — and a hollow success WARNS,
+# never dies: box is the host extra. Exercising it needs root + the network,
+# so grep the shipped script (repo precedent: the tag-refusal greps). Match
+# the CALL, not the word — the rationale comment says `command -v box` too.
+check "bootstrap: a box-install success is verified, not trusted" 0 "" \
+  grep -qE '^[[:space:]]*if command -v box' "$ROOT/commands/bootstrap.sh"
+check "bootstrap: a hollow box-install success warns, never dies" 0 "" \
+  grep -q "reported success but no 'box' is on PATH" "$ROOT/commands/bootstrap.sh"
+# Ordering: the effective check must sit AFTER the installer run it verifies.
+# Line-number compare, defaults fail closed (same idiom as the marker/install
+# ordering assert above; box_install_at is computed there).
+box_check_at="$(grep -nE '^[[:space:]]*if command -v box' "$ROOT/commands/bootstrap.sh" | head -n1 | cut -d: -f1)"
+check "bootstrap: the effective check follows the installer run" \
+  0 "" test "${box_install_at:-999999}" -lt "${box_check_at:-0}"
+# rig's delegation law caps the check's depth: rig never interrogates Incus —
+# the host verdict is box's own verb, and the "host set up" CLAIM is gated on
+# it. Two asserts: the gate exists as a call (not just prose naming the verb),
+# and the claim line sits inside/after it (line order, fail-closed defaults —
+# a claim that outruns its proof is exactly the overclaim this closes).
+check "bootstrap: the host-set-up claim is gated on box doctor" 0 "" \
+  grep -qE '^[[:space:]]*if box doctor' "$ROOT/commands/bootstrap.sh"
+doctor_at="$(grep -nE '^[[:space:]]*if box doctor' "$ROOT/commands/bootstrap.sh" | head -n1 | cut -d: -f1)"
+claim_at="$(grep -n 'box installed and host set up' "$ROOT/commands/bootstrap.sh" | head -n1 | cut -d: -f1)"
+check "bootstrap: the claim follows the doctor gate" \
+  0 "" test "${doctor_at:-999999}" -lt "${claim_at:-0}"
+check "bootstrap: a failed doctor warns without claiming the host" 0 "" \
+  grep -q "the CLI landed, the host stack is unproven" "$ROOT/commands/bootstrap.sh"
+# --- README: the box rename (#12) --------------------------------------------
+# The philosophy line must point at heavy-duty/box — the old claudebox slug
+# only works through a GitHub redirect that one squatted rename away from
+# breaking (box's own installer was already bitten by the rename once). A
+# negative grep (exit 1 = pass) keeps the stale slug from creeping back.
+check "README: no stale heavy-duty/claudebox links" 1 "" \
+  grep -n "heavy-duty/claudebox" "$ROOT/README.md"
+check "README: points at heavy-duty/box" 0 "" \
+  grep -q "github.com/heavy-duty/box" "$ROOT/README.md"
 if [ "$(id -u)" -ne 0 ]; then
   check "bootstrap: refuses non-root"      1 "must run as root" env TS_AUTHKEY=x "$ROOT/commands/bootstrap.sh" workload
   check "bootstrap: runner role parses, refuses non-root" 1 "must run as root" env TS_AUTHKEY=x "$ROOT/commands/bootstrap.sh" runner
@@ -170,6 +209,57 @@ else
   echo "skip: coolify backup non-root refusal (running as root)"
 fi
 
+# --- role-marker sanity: coolify verbs off the control plane (#25) -----------
+# Both coolify commands read /etc/rig/role and WARN — never die — when the
+# marker names a non-control-plane role: the likeliest story is the wrong SSH
+# session, but the marker is advisory and must not outrank the operator. The
+# warning fires BEFORE the root check (same testability rule as arg errors),
+# so a non-root run prints it and then hits the root refusal — provable here
+# with RIG_ROLE_MARKER pointed at fixtures (repo precedent: the close-root
+# marker gate). Counting fires proves silence too: a control-plane marker, an
+# absent marker, and a marker-less box must all stay quiet, because warning on
+# absence would nag every pre-marker box on every legitimate run.
+marker_warns() { # marker_warns <marker_path> <cmd...> — how many warnings fired
+  local marker="$1"; shift
+  env RIG_ROLE_MARKER="$marker" "$@" 2>&1 | grep -c "not a control-plane box" || true
+}
+MARKER_FIX="$(mktemp -d)"
+printf 'role=workload class=server host=no join=authkey\n'      > "$MARKER_FIX/workload"
+printf 'role=control-plane class=server host=no join=authkey\n' > "$MARKER_FIX/control-plane"
+printf 'role=control-plane\n'                                   > "$MARKER_FIX/bare-control-plane"
+if [ "$(id -u)" -ne 0 ]; then
+  check "coolify: warns on a non-control-plane marker" 0 "1" \
+    marker_warns "$MARKER_FIX/workload" "$ROOT/commands/coolify-install.sh" --version 4.1.2
+  check "coolify: control-plane marker stays silent" 0 "0" \
+    marker_warns "$MARKER_FIX/control-plane" "$ROOT/commands/coolify-install.sh" --version 4.1.2
+  # A bare marker line with no trailing traits must read the same as the full
+  # one — the guard must not couple to the marker's field formatting.
+  check "coolify: a bare 'role=control-plane' line (no traits) stays silent" 0 "0" \
+    marker_warns "$MARKER_FIX/bare-control-plane" "$ROOT/commands/coolify-install.sh" --version 4.1.2
+  check "coolify: absent marker stays silent (advisory, not a gate)" 0 "0" \
+    marker_warns "$MARKER_FIX/absent" "$ROOT/commands/coolify-install.sh" --version 4.1.2
+  # The warning must stay a warning: the run proceeds past it and stops at the
+  # root check (exit 1), never turned into a marker refusal.
+  check "coolify: the marker warns but never refuses" 1 "must run as root" \
+    env RIG_ROLE_MARKER="$MARKER_FIX/workload" "$ROOT/commands/coolify-install.sh" --version 4.1.2
+  check "coolify backup: warns on a non-control-plane marker" 0 "1" \
+    marker_warns "$MARKER_FIX/workload" "$ROOT/commands/coolify-backup-install.sh"
+  check "coolify backup: control-plane marker stays silent" 0 "0" \
+    marker_warns "$MARKER_FIX/control-plane" "$ROOT/commands/coolify-backup-install.sh"
+  check "coolify backup: the marker warns but never refuses" 1 "must run as root" \
+    env RIG_ROLE_MARKER="$MARKER_FIX/workload" "$ROOT/commands/coolify-backup-install.sh"
+else
+  echo "skip: coolify role-marker warning checks (running as root)"
+fi
+rm -rf "$MARKER_FIX"
+# Root runs skip the live checks above, so also pin the warning's presence in
+# both shipped scripts — a deleted advisory cannot ship green (repo precedent:
+# the staging/runner tag greps).
+check "coolify: marker warning present in the shipped script" 0 "" \
+  grep -q "not a control-plane box" "$ROOT/commands/coolify-install.sh"
+check "coolify backup: marker warning present in the shipped script" 0 "" \
+  grep -q "not a control-plane box" "$ROOT/commands/coolify-backup-install.sh"
+
 # --- rig db (ad-hoc dump/restore) -------------------------------------------
 check "bare db shows usage, exit 2"       2 "usage:" "$ROOT/bin/rig" db
 check "db --help exits 0"                 0 "usage:" "$ROOT/bin/rig" db --help