<!-- Generated from models/_canonical by scripts/build-model-variants.js. Edit the canonical source, not this file. Behavioural profile for GLM: scripts/model-profiles.json -->

#Task boundary

1. Implement only what the task names; no extra abstractions or files.
2. English-only comments and identifiers.
3. Stop when the checklist passes.

---

#Purpose

Rules for the response headers that constrain browser behaviour. Headers are cheap and deployable independently of application changes — but they are mitigations, not fixes. A strong CSP limits the damage of an XSS; it does not remove it.

---

#The set worth sending

css
Content-Security-Policy: default-src 'self'; script-src 'self' 'nonce-{RANDOM}' 'strict-dynamic'; object-src 'none'; base-uri 'none'; frame-ancestors 'none'
Strict-Transport-Security: max-age=31536000; includeSubDomains; preload
X-Content-Type-Options: nosniff
Referrer-Policy: strict-origin-when-cross-origin
Permissions-Policy: camera=(), microphone=(), geolocation=(), payment=()
Cross-Origin-Opener-Policy: same-origin
Cross-Origin-Resource-Policy: same-origin

#HSTS

max-age=31536000 (one year) with includeSubDomains.

Never add preload casually. Submission to the browser preload list is effectively irreversible on a useful timescale — every subdomain must serve HTTPS forever. Verify every subdomain first, including internal and legacy hosts.

Send HSTS only over HTTPS. A browser ignores it on a plaintext response, and sending it there suggests a misconfiguration.

#frame-ancestors over X-Frame-Options

frame-ancestors 'none' in CSP supersedes X-Frame-Options: DENY. Keep X-Frame-Options only for very old browsers; it takes no list of origins and its ALLOW-FROM value is not supported anywhere current.

---

#Deprecated — remove these

Copying a header block from an old article is how these persist. Check each against current browser support before shipping it.

---

#Why CSP is the one that matters

Of the headers above, Content-Security-Policy is the only one that changes what an attacker can achieve rather than merely what a browser reveals. The others close narrow gaps; CSP constrains script execution itself, which is why it is worth the deployment effort the rest do not require.

That effort is real. A strict policy will break inline scripts, inline styles and third-party widgets that were working, which is why the report-only phase below is not optional advice — it is how the policy gets deployed at all.

#Setting them

Set headers at one layer — the application, or the edge — not both. Duplicated and conflicting headers behave inconsistently across browsers and proxies.

js
import helmet from "helmet";

app.use(helmet({
  contentSecurityPolicy: {
    directives: {
      defaultSrc: ["'self'"],
      scriptSrc: ["'self'", (req, res) => `'nonce-${res.locals.nonce}'`, "'strict-dynamic'"],
      objectSrc: ["'none'"],
      baseUri: ["'none'"],
      frameAncestors: ["'none'"],
    },
  },
  hsts: { maxAge: 31536000, includeSubDomains: true },
  referrerPolicy: { policy: "strict-origin-when-cross-origin" },
}));

The nonce must be generated per response with a CSPRNG and never reused:

js
app.use((req, res, next) => {
  res.locals.nonce = crypto.randomBytes(16).toString("base64");
  next();
});

Never hard-code a nonce or derive it from anything predictable. A static nonce is equivalent to 'unsafe-inline'.

---

#Cookies

Cookie attributes are security headers by another name:

ini
Set-Cookie: sid=…; HttpOnly; Secure; SameSite=Lax; Path=/; Max-Age=1209600

Prefix a session cookie with __Host- where you can: the browser then enforces Secure, Path=/ and the absence of Domain, which prevents a subdomain from setting a cookie your application will trust.

---

#Caching sensitive responses

yaml
Cache-Control: no-store

Authenticated responses must not be cached by browsers or shared proxies. no-store is the correct directive; no-cache still permits storage with revalidation.

---

#Verifying

Test the deployed origin, not the configuration file — a proxy may add, strip or override headers:

bash
curl -sI https://app.example.com | grep -iE 'content-security|strict-transport|x-content-type|referrer|permissions'

Then check the report from a scanner such as Mozilla Observatory or securityheaders.com, and deploy CSP in Report-Only with a report-to endpoint before enforcing.

---

#Anti-patterns

---

#Checklist

• CSP set with a per-response CSPRNG nonce and strict-dynamic
• No 'unsafe-inline' or 'unsafe-eval' in script-src
• object-src 'none', base-uri 'none', frame-ancestors 'none' present
• HSTS max-age ≥ 1 year with includeSubDomains, HTTPS only
• preload used only after auditing every subdomain
• X-Content-Type-Options: nosniff on every response
• Referrer-Policy set to strict-origin-when-cross-origin or stricter
• Permissions-Policy denies unused device APIs
• X-XSS-Protection, Expect-CT and HPKP are absent
• Session cookies use HttpOnly, Secure, SameSite and __Host- where possible
• Authenticated responses send Cache-Control: no-store
• Headers are set at exactly one layer and verified against the live origin