perf(build): serve hashed assets from /_app with immutable caching

DEPLOY.md

@@ -15,7 +15,9 @@ against `api.github.com` with no server-side component.
 
 **Host the `dist/` build output on any static platform:** GitHub Pages, Netlify, Vercel,
 S3 + CloudFront, or any CDN that serves SPAs with `index.html` fallback for client-side
-routing.
+routing. Note: `public/.nojekyll` is copied into `dist/client/.nojekyll` during `pnpm run build`
+— keep this file when deploying to GitHub Pages so Jekyll does not strip the `_app/` directory
+containing hashed JS and CSS bundles.
 
 **What works without a backend:**
 - All dashboard features (Issues, PRs, Actions tabs)
@@ -214,6 +216,20 @@ The OAuth App access token is a permanent credential (no expiry). It is stored i
 
 The two tunnel endpoints (`/api/error-reporting` and `/api/csp-report`) receive untrusted browser data and forward it to Sentry. They are hardened with layered fail-fast guards (IP rate limit → Origin check → DSN validation). See `hack/docs/security-runbook.md` for the full threat model, per-endpoint rate limit values, Origin check behavior, CSP field sanitization details, Content-Length pre-check semantics, and fan-out amplification analysis.
 
+### Static asset caching
+
+- **Hashed bundles (`/_app/*`):** Vite outputs content-hashed JS and CSS bundles to `dist/client/_app/`. `public/_headers` configures `Cache-Control: public, max-age=31536000, immutable` (one year) on `/_app/*` while inheriting the global security headers from `/*`.
+- **Entry HTML and unhashed public assets (`index.html`, `/assets/*`):** `index.html` and unhashed images under `public/assets/*` (favicon, logos, social preview) have no custom `Cache-Control` rule in `public/_headers` and use the Cloudflare Workers Static Assets platform default (`Cache-Control: public, max-age=0, must-revalidate`), ensuring browsers revalidate `index.html` on every load.
+- **One-time cache miss on initial `/_app/` rollout:** Moving hashed bundles from `/assets/` to `/_app/` causes a one-time cache miss on the deployment that introduces `/_app/`.
+- **Stale-chunk 404 handling (`stale_asset_404`):** Under SPA `not_found_handling`, missing non-navigation requests (such as `<script>` or `import()` fetches for a chunk from a superseded deploy) invoke the Worker and `env.ASSETS.fetch(request)` returns the `200 text/html` SPA fallback. When `url.pathname.startsWith("/_app/")`, `assetResponse.status === 200`, and `Content-Type` starts with `text/html`, the Worker logs `stale_asset_404` and returns `404 Not Found` (`text/plain; charset=utf-8`) with `Cache-Control: no-store` and `SECURITY_HEADERS` so neither the browser nor CDN caches HTML under an immutable chunk URL.
+- **Accepted navigation edge case:** A direct browser *navigation* (`Sec-Fetch-Mode: navigate`) to a non-existent `/_app/*` URL is answered by Cloudflare's static asset router with `index.html` before the Worker runs and inherits the `/_app/*` immutable header from `_headers`. This is harmless (no application route or asset reference ever navigates to a missing `/_app/*` path) and avoids routing every `/_app/*` asset hit through the Worker.
+- **GitHub Pages `.nojekyll` requirement:** `public/.nojekyll` (emitted to `dist/client/.nojekyll`) disables Jekyll processing on GitHub Pages static-only deployments so the `_`-prefixed `_app/` directory is not dropped.
+- **Post-deploy verification:**
+  ```sh
+  curl -sI https://<domain>/_app/<hashed-file> | grep -i cache-control
+  # Expected: cache-control: public, max-age=31536000, immutable
+  ```
+
 ## Local Development
 
 Copy `.dev.vars.example` to `.dev.vars` and fill in your values. Wrangler picks up `.dev.vars` automatically for local `wrangler dev` runs.
@@ -262,6 +278,41 @@ Configure these rules in the Cloudflare dashboard under **Security → WAF**.
 
 ### Custom Rules
 
+#### Rule 1: Path Allowlist
+
+**Rule name:** Path Allowlist (block unknown paths)
+**Where:** Security → WAF → Custom Rules
+**Action:** Block
+
+**Previous expression (pre-`/_app/` rollout, synthesized from `scripts/waf-smoke-test.sh`):**
+```
+not (
+  http.request.uri.path in {"/" "/login" "/oauth/callback" "/onboarding" "/dashboard" "/settings" "/privacy" "/index.html"} or
+  starts_with(http.request.uri.path, "/assets/") or
+  starts_with(http.request.uri.path, "/api/")
+)
+```
+
+**Updated expression (allows `/_app/*` hashed chunks and static/metadata routes on `GET`/`HEAD`):**
+```
+not (
+  (
+    http.request.method in {"GET" "HEAD"} and (
+      http.request.uri.path in {"/" "/app" "/landing" "/landing.html" "/login" "/oauth/callback" "/onboarding" "/dashboard" "/settings" "/privacy" "/index.html" "/version.json" "/robots.txt" "/sitemap.xml" "/manifest.webmanifest" "/favicon.ico"} or
+      starts_with(http.request.uri.path, "/_app/") or
+      starts_with(http.request.uri.path, "/assets/")
+    )
+  ) or
+  starts_with(http.request.uri.path, "/api/")
+)
+```
+
+**Notes:**
+- Blocks vulnerability scanners and probes for unmapped paths (`/wp-admin`, `/.env`, `/.git/config`, arbitrary paths) at the edge before the Worker or static asset router runs.
+- `scripts/waf-smoke-test.sh` (`pnpm test:waf`) validates this rule before every commit (`prek`) and deploy (`deploy.yml`).
+
+#### Rule 3: Origin Gate
+
 **Rule name:** Block API requests without valid Origin
 **Where:** Security → WAF → Custom Rules
 **Expression:**

public/_headers

@@ -6,3 +6,6 @@
   Permissions-Policy: geolocation=(), microphone=(), camera=()
   Strict-Transport-Security: max-age=63072000; includeSubDomains; preload
   X-Frame-Options: DENY
+
+/_app/*
+  Cache-Control: public, max-age=31536000, immutable

scripts/waf-smoke-test.sh

@@ -6,7 +6,7 @@
 #   e.g. pnpm test:waf https://my-tracker.example.com
 #
 # Rules validated:
-#   1. Path Allowlist — blocks all paths except known SPA routes, /assets/*, /api/*
+#   1. Path Allowlist — blocks all paths except known SPA/static routes, /_app/*, /assets/*, /api/*, and public metadata files (/version.json, /robots.txt, /sitemap.xml, /manifest.webmanifest, /favicon.ico)
 #   2. Scanner User-Agents — challenges empty/malicious User-Agent strings
 #   3. Origin Gate — blocks /api/* requests without valid Origin header
 #   Rate limit rule exists but is not tested here (triggers a 10-minute IP block).
@@ -30,7 +30,7 @@ run_test() {
     [[ "$actual" != "000" && "$actual" != "502" && "$actual" != "503" ]] && break
     sleep "$attempt"
   done
-  if [[ "$actual" == "$expected" ]]; then
+  if [[ ",$expected," == *",$actual,"* ]]; then
     printf '  PASS  [%s] %s\n' "$actual" "$label"
   else
     printf '  FAIL  [%s] %s (expected %s)\n' "$actual" "$label" "$expected"
@@ -62,6 +62,7 @@ TESTS=(
   "200|GET /settings|${BASE}/settings"
   "200|GET /privacy|${BASE}/privacy"
   "307|GET /index.html (html_handling redirect)|${BASE}/index.html"
+  "403,200,404|GET /_app/nonexistent.js|${BASE}/_app/nonexistent.js"
   "200|GET /assets/nonexistent.js|${BASE}/assets/nonexistent.js"
   "200|GET /api/health (with Origin)|-H|Origin: ${BASE}|${BASE}/api/health"
   "400|POST /api/oauth/token (no body)|-X|POST|-H|Origin: ${BASE}|${BASE}/api/oauth/token"
@@ -92,9 +93,9 @@ TESTS=(
   "403|GET /eval-stdin.php|${BASE}/eval-stdin.php"
   "403|GET /.aws/credentials|${BASE}/.aws/credentials"
   "403|GET /.ssh/id_rsa|${BASE}/.ssh/id_rsa"
-  "403|GET /robots.txt|${BASE}/robots.txt"
-  "403|GET /sitemap.xml|${BASE}/sitemap.xml"
-  "403|GET /favicon.ico|${BASE}/favicon.ico"
+  "403,200|GET /robots.txt|${BASE}/robots.txt"
+  "403,200|GET /sitemap.xml|${BASE}/sitemap.xml"
+  "403,200|GET /favicon.ico|${BASE}/favicon.ico"
   "403|GET /random/garbage/path|${BASE}/random/garbage/path"
   # Rule 2: Scanner User-Agents — normal UAs
   "200|Normal browser UA|-H|User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36|${BASE}/"

src/worker/index.ts

@@ -107,7 +107,7 @@ function errorResponse(
   });
 }
 
-const SECURITY_HEADERS: Record<string, string> = {
+export const SECURITY_HEADERS: Record<string, string> = {
   "Strict-Transport-Security": "max-age=63072000; includeSubDomains; preload",
   "X-Content-Type-Options": "nosniff",
   "Referrer-Policy": "strict-origin-when-cross-origin",
@@ -1726,7 +1726,23 @@ export default Sentry.withSentry(
       }
 
       // Forward non-API requests to static assets
-      return env.ASSETS.fetch(request);
+      const assetResponse = await env.ASSETS.fetch(request);
+      if (
+        url.pathname.startsWith("/_app/") &&
+        assetResponse.status === 200 &&
+        (assetResponse.headers.get("content-type") ?? "").startsWith("text/html")
+      ) {
+        log("warn", "stale_asset_404", { pathname: url.pathname }, request);
+        return new Response(request.method === "HEAD" ? null : "Not Found", {
+          status: 404,
+          headers: {
+            ...SECURITY_HEADERS,
+            "Cache-Control": "no-store",
+            "Content-Type": "text/plain; charset=utf-8",
+          },
+        });
+      }
+      return assetResponse;
     },
   }
 );

tests/worker/static-assets.test.ts

@@ -0,0 +1,192 @@
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+import worker, { SECURITY_HEADERS, type Env } from "../../src/worker/index";
+import { collectLogs, findLog, ALLOWED_ORIGIN } from "./helpers";
+
+function makeEnv(overrides: Partial<Env> = {}): Env {
+  return {
+    ASSETS: { fetch: async () => new Response("asset") },
+    GITHUB_CLIENT_ID: "test_client_id",
+    GITHUB_CLIENT_SECRET: "test_client_secret",
+    ALLOWED_ORIGIN,
+    SENTRY_DSN: "https://abc123@o123456.ingest.sentry.io/7890123",
+    SESSION_KEY: "dGVzdC1zZXNzaW9uLWtleQ==",
+    SEAL_KEY: "dGVzdC1zZWFsLWtleQ==",
+    TURNSTILE_SECRET_KEY: "test-turnstile-secret",
+    PROXY_RATE_LIMITER: { limit: vi.fn().mockResolvedValue({ success: true }) },
+    ...overrides,
+  };
+}
+
+describe("Worker static asset handling for /_app/*", () => {
+  let consoleSpy: {
+    info: ReturnType<typeof vi.spyOn>;
+    warn: ReturnType<typeof vi.spyOn>;
+    error: ReturnType<typeof vi.spyOn>;
+  };
+
+  beforeEach(() => {
+    consoleSpy = {
+      info: vi.spyOn(console, "info").mockImplementation(() => {}),
+      warn: vi.spyOn(console, "warn").mockImplementation(() => {}),
+      error: vi.spyOn(console, "error").mockImplementation(() => {}),
+    };
+  });
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  it("returns 404 with no-store, SECURITY_HEADERS, and logs stale_asset_404 on GET /_app/* when ASSETS returns 200 text/html", async () => {
+    const env = makeEnv({
+      ASSETS: {
+        fetch: async () =>
+          new Response("<!doctype html><html><body>SPA</body></html>", {
+            status: 200,
+            headers: { "Content-Type": "text/html" },
+          }),
+      },
+    });
+    const req = new Request("https://gh.gordoncode.dev/_app/missing-abc123.js", {
+      method: "GET",
+    });
+
+    const resp = await worker.fetch(req, env);
+
+    expect(resp.status).toBe(404);
+    expect(await resp.text()).toBe("Not Found");
+    expect(resp.headers.get("Cache-Control")).toBe("no-store");
+    expect(resp.headers.get("Content-Type")).toBe("text/plain; charset=utf-8");
+    for (const [key, value] of Object.entries(SECURITY_HEADERS)) {
+      expect(resp.headers.get(key)).toBe(value);
+    }
+
+    const logs = collectLogs(consoleSpy);
+    const staleLog = findLog(logs, "stale_asset_404");
+    expect(staleLog).toBeDefined();
+    expect(staleLog?.level).toBe("warn");
+    expect(staleLog?.entry.pathname).toBe("/_app/missing-abc123.js");
+  });
+
+  it("returns 404 on GET /_app/* when ASSETS returns 200 text/html; charset=utf-8", async () => {
+    const env = makeEnv({
+      ASSETS: {
+        fetch: async () =>
+          new Response("<!doctype html><html><body>SPA</body></html>", {
+            status: 200,
+            headers: { "Content-Type": "text/html; charset=utf-8" },
+          }),
+      },
+    });
+    const req = new Request("https://gh.gordoncode.dev/_app/missing-abc123.js", {
+      method: "GET",
+    });
+
+    const resp = await worker.fetch(req, env);
+
+    expect(resp.status).toBe(404);
+    expect(await resp.text()).toBe("Not Found");
+    expect(resp.headers.get("Cache-Control")).toBe("no-store");
+    expect(resp.headers.get("Content-Type")).toBe("text/plain; charset=utf-8");
+    for (const [key, value] of Object.entries(SECURITY_HEADERS)) {
+      expect(resp.headers.get(key)).toBe(value);
+    }
+  });
+
+  it("returns 404 with empty body on HEAD /_app/* when ASSETS returns 200 text/html", async () => {
+    const env = makeEnv({
+      ASSETS: {
+        fetch: async () =>
+          new Response("<!doctype html><html><body>SPA</body></html>", {
+            status: 200,
+            headers: { "Content-Type": "text/html" },
+          }),
+      },
+    });
+    const req = new Request("https://gh.gordoncode.dev/_app/missing-abc123.js", {
+      method: "HEAD",
+    });
+
+    const resp = await worker.fetch(req, env);
+
+    expect(resp.status).toBe(404);
+    expect(resp.body).toBeNull();
+    expect(await resp.text()).toBe("");
+    expect(resp.headers.get("Cache-Control")).toBe("no-store");
+    expect(resp.headers.get("Content-Type")).toBe("text/plain; charset=utf-8");
+    for (const [key, value] of Object.entries(SECURITY_HEADERS)) {
+      expect(resp.headers.get(key)).toBe(value);
+    }
+  });
+
+  it("passes through 200 application/javascript response for /_app/index-abc.js", async () => {
+    const env = makeEnv({
+      ASSETS: {
+        fetch: async () =>
+          new Response("console.log('ok');", {
+            status: 200,
+            headers: {
+              "Content-Type": "application/javascript",
+              "Cache-Control": "public, max-age=31536000, immutable",
+            },
+          }),
+      },
+    });
+    const req = new Request("https://gh.gordoncode.dev/_app/index-abc.js", {
+      method: "GET",
+    });
+
+    const resp = await worker.fetch(req, env);
+
+    expect(resp.status).toBe(200);
+    expect(await resp.text()).toBe("console.log('ok');");
+    expect(resp.headers.get("Content-Type")).toBe("application/javascript");
+    expect(resp.headers.get("Cache-Control")).toBe(
+      "public, max-age=31536000, immutable"
+    );
+    expect(findLog(collectLogs(consoleSpy), "stale_asset_404")).toBeUndefined();
+  });
+
+  it("passes through 200 text/html SPA fallback for non-/_app/ routes like /dashboard", async () => {
+    const html = "<!doctype html><html><body><div id='root'></div></body></html>";
+    const env = makeEnv({
+      ASSETS: {
+        fetch: async () =>
+          new Response(html, {
+            status: 200,
+            headers: { "Content-Type": "text/html; charset=utf-8" },
+          }),
+      },
+    });
+    const req = new Request("https://gh.gordoncode.dev/dashboard", {
+      method: "GET",
+    });
+
+    const resp = await worker.fetch(req, env);
+
+    expect(resp.status).toBe(200);
+    expect(await resp.text()).toBe(html);
+    expect(resp.headers.get("Content-Type")).toBe("text/html; charset=utf-8");
+    expect(findLog(collectLogs(consoleSpy), "stale_asset_404")).toBeUndefined();
+  });
+
+  it("passes through 304 Not Modified response from ASSETS for /_app/x.js", async () => {
+    const env = makeEnv({
+      ASSETS: {
+        fetch: async () =>
+          new Response(null, {
+            status: 304,
+            headers: { ETag: '"abc123"' },
+          }),
+      },
+    });
+    const req = new Request("https://gh.gordoncode.dev/_app/x.js", {
+      method: "GET",
+    });
+
+    const resp = await worker.fetch(req, env);
+
+    expect(resp.status).toBe(304);
+    expect(resp.headers.get("ETag")).toBe('"abc123"');
+    expect(findLog(collectLogs(consoleSpy), "stale_asset_404")).toBeUndefined();
+  });
+});

vite.config.ts

@@ -8,4 +8,11 @@ export default defineConfig({
   build: {
     outDir: "dist",
   },
+  environments: {
+    client: {
+      build: {
+        assetsDir: "_app",
+      },
+    },
+  },
 });