diff --git a/scripts/pre-pr-checks.sh b/scripts/pre-pr-checks.sh
new file mode 100755
index 000000000..f18190d6e
--- /dev/null
+++ b/scripts/pre-pr-checks.sh
@@ -0,0 +1,120 @@
+#!/usr/bin/env bash
+# Copyright 2026 bong-water-water-bong
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+# pre-pr-checks.sh [--base REF] [--list]
+#
+# `docs/pre-pr-checks.md` (engine#2311) lists six host-side checks to run before asking for a merge.
+# Asking a lane to type six commands, two of which need an argument, is how one gets skipped: three
+# stale-baseline incidents in two days, each a check that exists and was not run.
+#
+# This runs all six and prints ONE line per check, so the answer is per-check rather than "the paste
+# looked empty". A check that cannot run says `skip` and why -- a skipped check is not a pass, and the
+# summary keeps the count so a green line cannot hide one (the reason `tools/ci/tmp_path_check.py`
+# prints its excluded files instead of excluding them silently).
+#
+# Deliberate: a NEW file must be `git add`ed first. Three of these six read the index or the tracked
+# tree, so an unadded file is invisible to them and the run reads green -- the same trap that made
+# `tools/copyright.py --check` skip untracked files by design.
+set -uo pipefail
+
+# usage(): the header comment as prose, anchored on the comment block rather than a line range -- two
+# scripts in this repo drifted past the end of their header and printed code to whoever asked for
+# help (engine#2129).
+usage() { awk 'NR>1 && /^#/ {sub(/^# ?/,""); print; next} NR>1 {exit}' "$0"; }
+
+base=origin/main
+list_only=0
+while [ $# -gt 0 ]; do
+    case "$1" in
+        --base) base=${2:-}; shift 2 || { usage >&2; exit 2; } ;;
+        --base=*) base=${1#--base=}; shift ;;
+        --list) list_only=1; shift ;;
+        -h|--help) usage; exit 0 ;;
+        *) echo "pre-pr-checks: unknown argument '$1'" >&2; usage >&2; exit 2 ;;
+    esac
+done
+[ -n "$base" ] || { echo "pre-pr-checks: --base needs a ref" >&2; exit 2; }
+
+# Run from the toplevel: the six commands name `tools/...` and `config/...` relative to it, so a caller
+# in scripts/ got twelve `can't open file` lines and rc 1 with every row red for the wrong reason.
+top=$(git rev-parse --show-toplevel 2>/dev/null) || { echo "pre-pr-checks: not inside a git work tree" >&2; exit 2; }
+cd "$top" || { echo "pre-pr-checks: cannot enter $top" >&2; exit 2; }
+
+# A `--base` that does not resolve is a USAGE error, not a skip: the caller asked for a comparison that
+# cannot happen, and a run that prints "all 5 check(s) green, 1 skipped" then exits 0 is the shape this
+# script exists to stop (review of engine#2313). A skip is for a check that cannot run in this
+# environment; an argument that cannot be resolved is the caller's to fix.
+git rev-parse --verify -q "$base^{commit}" >/dev/null || {
+    echo "pre-pr-checks: --base '$base' does not resolve to a commit here" >&2
+    echo "  the registration check compares against it; pass a ref that exists (default: origin/main)" >&2
+    exit 2
+}
+
+# The checks, in the order docs/pre-pr-checks.md lists them: cheapest and most common first, the two
+# that need a revision last.
+checks=(
+    "tools/copyright.py --check"
+    "tools/ci/tests_referenced_check.py"
+    "tools/ci/ci_targets_check.py --check"
+    "tools/ci/tmp_path_check.py --check"
+    "tools/ci/submodule_pin_check.py --check"
+    "tools/ci/registration_loss_check.py --check --base $base"
+)
+
+if [ "$list_only" = 1 ]; then
+    for c in "${checks[@]}"; do echo "python3 $c"; done
+    exit 0
+fi
+
+ok=0; failed=0; skipped=0; detail=""
+for c in "${checks[@]}"; do
+    name=${c%% *}
+    out=$(python3 $c 2>&1); rc=$?
+    last=$(printf '%s\n' "$out" | grep -v '^$' | tail -1)
+    if [ $rc -eq 0 ]; then
+        # `submodule_pin_check --check` exits 0 having compared NOTHING when submodules are declared but
+        # none is checked out -- `0 submodule(s) at their pinned commit, N not initialised` -- which would
+        # print `ok` for a check that ran no comparison, the shape this script exists to stop. Matched on
+        # the tool's own summary rather than re-deriving the condition, because the tool is the thing that
+        # knows. Its OTHER "nothing to compare" case, `no submodules reported - cannot tell, and that is not
+        # a pass`, exits 2 and is left to fail: the tool already refuses to call that a pass.
+        # `--require-initialised` is where a release build turns this one into a failure.
+        case "$last" in
+            "submodule_pin: 0 submodule(s) at their pinned commit, "*)
+                printf 'skip  %-42s nothing compared: %s\n' "$name" "${last#submodule_pin: }"
+                skipped=$((skipped + 1)); continue ;;
+        esac
+        printf 'ok    %-42s %s\n' "$name" "$last"
+        ok=$((ok + 1))
+    else
+        printf 'FAIL  %-42s %s\n' "$name" "$last"
+        failed=$((failed + 1))
+        # The whole output, not its last line: the case that matters is a checker whose summary is
+        # generic and whose detail line names the commit or file (`submodule_pin_check` prints "NOT the
+        # commit this tree pins" above a summary that only counts), and the last line sends the reader
+        # to the wrong place.
+        detail="$detail$name:"$'\n'"$out"$'\n'
+    fi
+done
+
+echo
+if [ $failed -gt 0 ]; then
+    echo "pre-pr-checks: $failed failed, $ok ok, $skipped skipped"
+    printf '%s' "$detail"
+    exit 1
+fi
+echo "pre-pr-checks: all $ok check(s) green, $skipped skipped"
+exit 0