feat(doctor): brain-home fleet-state check (#1298 follow-up) (#1301)

Co-authored-by: Jason Woltje <[email protected]>

packages/mosaic/framework/tools/_scripts/mosaic-doctor

@@ -255,6 +255,68 @@ fleet_declared_transport() {
   printf '%s\n' "${declared:-tmux}"
 }
 
+# Brain-home fleet-state resolution (#1298; canon STRUCTURE-CANON §2).
+#
+# Seat launch envs, roles.local overrides, and profile working copies resolve
+# from the brain home when one is active; roster, baseline roles, run/, and
+# services stay under MOSAIC_HOME. This check surfaces which tree fleet state
+# resolves from and the drift a launch would otherwise hit at runtime:
+#
+# - a stale MOSAIC_BRAIN_HOME pointing at a directory with no fleet/agents is a
+#   misconfiguration the resolver honors (explicit wins) — warn, don't pass;
+# - a symlinked brain or agents dir defeats the managed-directory boundary;
+# - a group/world-readable agents dir violates the 0700 projection boundary;
+# - env files left in the config-home tree while a brain is active are split
+#   state — the write path rejects NEW split writes, but nothing would ever
+#   tell the operator the old files are stranded.
+resolve_brain_home() {
+  local explicit="${MOSAIC_BRAIN_HOME:-}"
+  if [[ -n "$(printf '%s' "$explicit" | tr -d '[:space:]')" ]]; then
+    printf '%s' "$explicit"
+    return
+  fi
+  if [[ "$(cd "$MOSAIC_HOME" 2>/dev/null && pwd -P)" == "$HOME/.config/mosaic" \
+     && -d "$HOME/.mosaic/fleet/agents" ]]; then
+    printf '%s' "$HOME/.mosaic"
+    return
+  fi
+  printf '%s' "$MOSAIC_HOME"
+}
+
+check_brain_home() {
+  local brain agents mode
+  brain="$(resolve_brain_home)"
+
+  if [[ "$brain" == "$MOSAIC_HOME" ]]; then
+    pass "Fleet state home: $MOSAIC_HOME (legacy single-tree; no brain adopted)"
+    return
+  fi
+
+  agents="$brain/fleet/agents"
+  if [[ ! -d "$agents" ]]; then
+    warn "Brain home '$brain' has no fleet/agents — seat envs will not resolve from it. Point MOSAIC_BRAIN_HOME at a brain carrying fleet/agents, or unset it."
+    return
+  fi
+  if [[ -L "$brain" || -L "$agents" ]]; then
+    warn "Brain fleet-state path resolves through a symlink ($brain) — the managed-directory boundary requires regular directories."
+    return
+  fi
+
+  mode="$(stat -c '%a' -- "$agents" 2>/dev/null)" || mode=""
+  if [[ -n "$mode" ]] && (( (8#$mode & 8#077) != 0 )); then
+    warn "Brain agents dir '$agents' is group/world-accessible (mode $mode) — the projection boundary requires 0700."
+    return
+  fi
+
+  if [[ -d "$MOSAIC_HOME/fleet/agents" ]] \
+     && ls "$MOSAIC_HOME/fleet/agents/"*.env* >/dev/null 2>&1; then
+    warn "Fleet env files exist in BOTH trees — brain '$brain' is active but '$MOSAIC_HOME/fleet/agents' still carries env files (split state). Migrate them (mosaic fleet regen) and remove the config-home copies."
+    return
+  fi
+
+  pass "Fleet state home: $brain (brain active); roster + templates: $MOSAIC_HOME"
+}
+
 check_fleet_transport() {
   local transport
   transport="$(fleet_declared_transport)"
@@ -273,6 +335,8 @@ check_fleet_transport() {
 
 check_fleet_transport
 
+check_brain_home
+
 # Legacy migration surfaces should no longer contain symlink trees.
 legacy_paths=(
   "$HOME/.claude/agent-guides"

packages/mosaic/framework/tools/_scripts/test-brain-home-check.sh

@@ -0,0 +1,108 @@
+#!/usr/bin/env bash
+# Covers the brain-home fleet-state check in `mosaic-doctor` (#1298 follow-up).
+#
+# The functions are extracted from the shipped script rather than copied here
+# (same discipline as test-fleet-transport-check.sh): a test that carries its
+# own copy of the logic keeps passing after the shipped copy changes.
+# Extraction is by exact function header and a closing brace in column one.
+set -euo pipefail
+
+SCRIPT_DIR=$(cd -- "$(dirname "$0")" && pwd)
+DOCTOR="$SCRIPT_DIR/mosaic-doctor"
+
+fail() {
+  echo "FAIL: $*" >&2
+  exit 1
+}
+
+[ -f "$DOCTOR" ] || fail "missing mosaic-doctor at $DOCTOR"
+
+extract_function() {
+  local name="$1"
+  local extracted
+  extracted=$(sed -n "/^${name}() {/,/^}/p" "$DOCTOR")
+  [ -n "$extracted" ] || fail "could not extract ${name}() from mosaic-doctor — script reshaped?"
+  printf '%s\n' "$extracted"
+}
+
+for fn in resolve_brain_home check_brain_home; do
+  extract_function "$fn" >/dev/null
+done
+
+warn_count=0
+warn() { warn_count=$((warn_count + 1)); echo "[WARN] $*"; }
+pass() { echo "[OK] $*"; return 0; }
+
+eval "$(extract_function resolve_brain_home)"
+eval "$(extract_function check_brain_home)"
+
+ROOT=$(mktemp -d)
+trap 'rm -rf "$ROOT"' EXIT
+
+run_case() {
+  # label, expect (ok|warn), then env assignments as arguments.
+  # The check runs under `env` in a subshell, so its warn() also prints a
+  # sentinel the parent counts — a subshell counter would never be visible.
+  local label="$1" expect="$2"
+  shift 2
+  local out warns
+  out=$(env "$@" bash -c "warn() { echo \"[WARN] \$*\"; }; pass() { echo \"[OK] \$*\"; return 0; }; $(extract_function resolve_brain_home); $(extract_function check_brain_home); check_brain_home" 2>&1)
+  warns=$(printf '%s\n' "$out" | grep -c '^\[WARN\]' || true)
+  if [[ "$expect" == ok && "$warns" -eq 0 ]]; then
+    echo "ok - $label"
+  elif [[ "$expect" == warn && "$warns" -gt 0 ]]; then
+    echo "ok - $label (warned)"
+  else
+    echo "output: $out" >&2
+    fail "$label: expected $expect (warns=$warns)"
+  fi
+}
+
+# ── legacy: no brain, custom home never adopts ─────────────────────────────
+mkdir -p "$ROOT/legacy-mosaic/fleet/agents"
+run_case "custom home without brain stays legacy" ok \
+  MOSAIC_HOME="$ROOT/legacy-mosaic" HOME="$ROOT"
+
+# ── healthy brain at the default config home ───────────────────────────────
+mkdir -p "$ROOT/home/.config/mosaic" "$ROOT/home/.mosaic/fleet/agents"
+chmod 700 "$ROOT/home/.mosaic/fleet/agents"
+run_case "default home adopts healthy brain" ok \
+  MOSAIC_HOME="$ROOT/home/.config/mosaic" HOME="$ROOT/home"
+
+# ── explicit MOSAIC_BRAIN_HOME to a brain without fleet/agents → warn ──────
+mkdir -p "$ROOT/brain-noagents/fleet" "$ROOT/config"
+run_case "explicit brain without agents warns" warn \
+  MOSAIC_HOME="$ROOT/config" HOME="$ROOT" MOSAIC_BRAIN_HOME="$ROOT/brain-noagents"
+
+# ── explicit MOSAIC_BRAIN_HOME to a healthy brain → ok ─────────────────────
+mkdir -p "$ROOT/brain-ok/fleet/agents" "$ROOT/config2"
+chmod 700 "$ROOT/brain-ok/fleet/agents"
+run_case "explicit healthy brain passes" ok \
+  MOSAIC_HOME="$ROOT/config2" HOME="$ROOT" MOSAIC_BRAIN_HOME="$ROOT/brain-ok"
+
+# ── group-readable agents dir → warn (0700 boundary) ───────────────────────
+mkdir -p "$ROOT/brain-loose/fleet/agents" "$ROOT/config3"
+chmod 750 "$ROOT/brain-loose/fleet/agents"
+run_case "group-readable brain agents warns" warn \
+  MOSAIC_HOME="$ROOT/config3" HOME="$ROOT" MOSAIC_BRAIN_HOME="$ROOT/brain-loose"
+
+# ── symlinked agents dir → warn (managed-directory boundary) ───────────────
+mkdir -p "$ROOT/brain-link/real-agents" "$ROOT/brain-link/fleet" "$ROOT/config4"
+ln -s "$ROOT/brain-link/real-agents" "$ROOT/brain-link/fleet/agents"
+run_case "symlinked brain agents warns" warn \
+  MOSAIC_HOME="$ROOT/config4" HOME="$ROOT" MOSAIC_BRAIN_HOME="$ROOT/brain-link"
+
+# ── split state: envs in BOTH trees → warn ─────────────────────────────────
+mkdir -p "$ROOT/brain-split/fleet/agents" "$ROOT/config5/fleet/agents"
+chmod 700 "$ROOT/brain-split/fleet/agents" "$ROOT/config5/fleet/agents"
+touch "$ROOT/config5/fleet/agents/coder0.env.generated"
+run_case "env files in both trees warns (split state)" warn \
+  MOSAIC_HOME="$ROOT/config5" HOME="$ROOT" MOSAIC_BRAIN_HOME="$ROOT/brain-split"
+
+# ── config-home agents dir WITHOUT env files alongside a brain → ok ────────
+mkdir -p "$ROOT/brain-clean/fleet/agents" "$ROOT/config6/fleet/agents"
+chmod 700 "$ROOT/brain-clean/fleet/agents" "$ROOT/config6/fleet/agents"
+run_case "empty config-home agents dir alongside brain passes" ok \
+  MOSAIC_HOME="$ROOT/config6" HOME="$ROOT" MOSAIC_BRAIN_HOME="$ROOT/brain-clean"
+
+echo "ok - mosaic-doctor brain-home check"