diff --git a/.gitea/workflows/deploy.yml b/.gitea/workflows/deploy.yml
index 414a3b4..4fe4c42 100644
--- a/.gitea/workflows/deploy.yml
+++ b/.gitea/workflows/deploy.yml
@@ -88,3 +88,28 @@ jobs:
           docker ps --filter name=recurring-checklist --format '{{.Names}} {{.Status}}'
           docker logs --tail 30 recurring-checklist
           exit 1
+
+      - name: Tell Christian the deploy broke
+        # The web deploy had no failure notification at all until now, which is the same
+        # silent-failure shape as the backup: it either worked or you found out by looking.
+        #
+        # NOT covered by this: a runner that wedges. On 2026-09-24 the NAS runner claimed
+        # this job, ran none of its six steps, and Gitea timed it out — no step ran, so no
+        # step could report. Catching that needs something watching from outside the run.
+        if: failure()
+        run: |
+          D=/volume2/docker/recurring-checklist
+          BODY_FILE="$(mktemp)"
+          cat > "$BODY_FILE" <<MSG
+          A push to main failed to deploy the website to the NAS.
+
+          The running container was left alone — the deploy recreates it in place and only
+          after the files are copied, so the previous version should still be serving. Worth
+          confirming: https://check.watersmedia.org
+
+          Full log:
+          http://chronicles-ab.waterschurch.org:3000/christian/recurring-checklist/actions
+          MSG
+          CHECKLIST_ENV_FILE="$D/.env" CHECKLIST_NODE_MODULES="$D/node_modules" \
+            node scripts/alert.js "Website deploy FAILED" "$(cat "$BODY_FILE")" || true
+          rm -f "$BODY_FILE"
diff --git a/scripts/alert.js b/scripts/alert.js
index 642eb92..a926e69 100755
--- a/scripts/alert.js
+++ b/scripts/alert.js
@@ -15,7 +15,11 @@ const fs = require('fs');
 const path = require('path');
 
 const ROOT = path.resolve(__dirname, '..');
-const ENV_FILE = path.join(ROOT, 'server', '.env');
+// Overridable because the deploy job runs inside a container that has the repo checked out
+// but no server/.env and no node_modules — those live on the NAS at the mounted deploy
+// directory instead. Without this the deploy could not report its own failures.
+const ENV_FILE = process.env.CHECKLIST_ENV_FILE || path.join(ROOT, 'server', '.env');
+const NODE_MODULES = process.env.CHECKLIST_NODE_MODULES || path.join(ROOT, 'server', 'node_modules');
 
 // Minimal .env reader — the server uses dotenv, but this script has to run from a plain
 // LaunchAgent with no cwd guarantees and no node_modules of its own.
@@ -56,8 +60,8 @@ async function main() {
 
   // nodemailer lives in server/node_modules; this script has none of its own.
   let nodemailer;
-  try { nodemailer = require(path.join(ROOT, 'server', 'node_modules', 'nodemailer')); }
-  catch { console.error('alert.js: nodemailer not installed — run `npm install` in server/'); process.exit(1); }
+  try { nodemailer = require(path.join(NODE_MODULES, 'nodemailer')); }
+  catch { console.error(`alert.js: nodemailer not found under ${NODE_MODULES} — run \`npm install\` in server/`); process.exit(1); }
 
   const transport = nodemailer.createTransport({
     host: env.SMTP_HOST,