Without logging in or testing anything aggressive, check the public security settings of the site I work on (HTTPS and security headers). Explain in plain language what is missing and why it matters, before suggesting changes.
Check a site’s public security settings
You want the common, preventable security gaps found before someone else finds them, without any risky testing.
- Any
- Claude Code
- Grok Build
- Cursor
- Codex CLI
For: Owner, Office manager, Tech person, Anyone at work · Updated November 10, 2026
A starting point: fill in the [brackets] with your own details, and check the answer before you use it. Tool screens change often, so yours may look a little different.
The prompts
Copy a quick prompt, or grab the full skill below.
List the outside scripts the site I work on loads, like analytics, chat, or fonts. For each, say what it does and whether it looks unused. Report only; don't remove anything.
Here are the security headers from the site I work on: [paste]. Explain what each one does, which are missing, and which matter most, in plain English. Don't change anything yet.What this covers: A passive outside-in review of HTTPS, security headers, cookies, forms and spam protection, exposed files, the admin login, and third-party scripts.
For anyone improving a site or app, whether a small-business owner or someone at a larger company practicing better prompts. Most website break-ins use common, preventable gaps, and checking for them from the outside is safe and quick. Paste a prompt into any AI assistant that can see the site or repo.
Two ways to use it
- Any AI chat (Copilot, ChatGPT, Claude, Grok): copy one of the quick prompts above. Each one asks for a plain-language report first, so nothing changes until you say so.
- A coding agent working in the site’s code (Claude Code, Grok Build, Cursor, Codex CLI): save the full skill below as
SKILL.mdin a folder namedsecurity-hardening-audit, for example.claude/skills/security-hardening-audit/SKILL.md(Grok Build also reads that location, or use.grok/skills/). Then ask the agent to use thesecurity-hardening-auditskill. It audits first, makes the smallest safe changes in a review branch, and ends with a report.
Download SKILL.md · All Website Care skills
The full skill (SKILL.md)
Last reviewed Nov 10, 2026. Standards and tools change, so check anything dated before relying on it.
---
name: security-hardening-audit
description: Use when asked to check or harden a website's or web app's security posture from the outside in - HTTPS/HSTS, security headers and CSP, cookies, forms (CSRF, spam, bot protection), exposed files (.env, .git, backups, debug pages), admin and login surface, and third-party scripts - and to propose fixes as a review PR. Audit is read-only and non-intrusive by default. Not a penetration test, not for dependency updates (use dependency-update-pass).
---
# Security hardening audit
Last reviewed: 2026-11-10
## Quick prompts
For anyone improving a site or app, whether a small-business owner or someone at a larger company practicing better prompts. Most website break-ins use common, preventable gaps, and checking for them from the outside is safe and quick. Paste a prompt into any AI assistant that can see the site or repo.
1. "Without logging in or testing anything aggressive, check the public security settings of the site I work on (HTTPS and security headers). Explain in plain language what is missing and why it matters, before suggesting changes."
*Why this prompt works:* Set limits up front (read-only, nothing aggressive) so the assistant stays safe.
2. "List the outside scripts the site I work on loads, like analytics, chat, or fonts. For each, say what it does and whether it looks unused. Report only; don't remove anything."
*Why this prompt works:* Ask for a report first: removing a script can break something someone relies on.
3. "Here are the security headers from the site I work on: [paste]. Explain what each one does, which are missing, and which matter most, in plain English. Don't change anything yet."
*Why this prompt works:* Giving the assistant the actual data beats asking it to guess.
Find the common, preventable weaknesses of a public website or web app and propose the smallest safe fixes. This is a passive, owner-authorized review, not an attack.
## Hard limits
- Only test sites the owner controls and has asked you to check. No scanning of third parties.
- Passive and polite: normal requests, a browser user agent, a pause between requests, no brute force, no fuzzing, no load tests, no exploit attempts, no form submissions with real data.
- If you find an exposed secret (key in a JS bundle, `.env` served publicly), **do not print, copy, test, or commit it**. Report the URL and type only, and tell the owner to rotate it.
- Fixes go in a review PR. Header and config changes at the CDN/host, DNS, mail (SPF/DKIM/DMARC), and secret rotation need the owner's explicit approval and are usually done by them.
- Never weaken an existing control to make something work.
## Inputs to gather
1. Domains and subdomains in scope (production, staging, preview).
2. Hosting/CDN and where headers are set (framework middleware, `_headers` file, server config, CDN rules).
3. Stack: static, SSR, SPA plus API, CMS (WordPress etc.), auth provider.
4. Forms and endpoints that accept input; how spam is handled today.
5. Third-party scripts and why each exists (analytics, ads, chat, embeds, fonts).
## Step 1: Classify
| Type | Focus | Lower priority |
|---|---|---|
| Static content site (articles, marketing, docs) | HTTPS/HSTS, headers, CSP for third-party scripts, exposed files, form spam, staging exposure | Session handling (usually none) |
| CMS site | Admin path exposure, login rate limiting, plugin/theme versions shown, XML-RPC, user enumeration, backups in web root | — |
| E-commerce | Payment page isolation (hosted fields/iframes), CSP on checkout, cookie flags, third-party script inventory (skimming risk) | — |
| SaaS / app with auth | Cookies (`Secure`, `HttpOnly`, `SameSite`), CSRF on state-changing requests, CORS, session timeout, password reset flow, MFA availability, error pages leaking stack traces | — |
| SPA + API | CORS allowlist, tokens kept out of `localStorage` where possible, API error verbosity, source maps in production | — |
## Step 2: Check (read-only)
**Transport**
- HTTP redirects to HTTPS with one hop; certificate valid and auto-renewing.
- `Strict-Transport-Security: max-age=31536000; includeSubDomains` once every subdomain is HTTPS. Only add `preload` after the owner understands it is hard to undo.
**Headers** (compare against the OWASP Secure Headers recommendations)
- `Content-Security-Policy`: start as `Content-Security-Policy-Report-Only`, list every real source, avoid `unsafe-inline`/`unsafe-eval` where feasible (nonces or hashes instead), include `frame-ancestors`, `base-uri 'self'`, `object-src 'none'`, `form-action`.
- `X-Content-Type-Options: nosniff`; `Referrer-Policy: strict-origin-when-cross-origin`; `Permissions-Policy` turning off unused features; `frame-ancestors` (or `X-Frame-Options: DENY/SAMEORIGIN` for old browsers).
- Remove version-leaking headers (`X-Powered-By`, detailed `Server`).
- `Cross-Origin-Opener-Policy: same-origin` where popups do not need it.
**Cookies**: every session or auth cookie has `Secure`, `HttpOnly`, and `SameSite=Lax` or `Strict`; consider the `__Host-` prefix.
**Forms and input**
- CSRF protection on state-changing requests (token or SameSite plus origin check).
- Spam/bot protection that does not hurt accessibility: honeypot field, server-side validation, rate limiting, and a privacy-friendly challenge (for example Cloudflare Turnstile or similar) verified **server-side**.
- Contact forms must not allow header injection or arbitrary recipients; success and error pages should not echo raw input.
**Exposed files and surfaces** (request with HEAD/GET, expect 403/404)
- `/.env`, `/.git/HEAD`, `/.git/config`, `/.DS_Store`, `/backup.zip`, `*.sql`, `*.bak`, `/wp-config.php.bak`, `/phpinfo.php`, `/server-status`, `/.well-known/` contents, production source maps (`*.map`) if they reveal private code.
- Directory listing off. Custom 404/500 pages that do not show stack traces.
- Admin/login paths: rate limited, MFA available, not indexed. Staging/preview: password- or access-protected, or at least `noindex`.
- `/.well-known/security.txt` with a contact (RFC 9116).
**Third-party scripts**: inventory each one, its purpose, and whether it is still used. Use Subresource Integrity (`integrity` + `crossorigin`) for static CDN files; remove dead tags.
**Email domain (report only)**: SPF, DKIM, DMARC records exist. Changing them is a DNS change that needs approval.
Helpful tools: browser DevTools network panel, `curl -sI` for headers on hosts that allow it, Mozilla HTTP Observatory, securityheaders.com, SSL Labs. Note that third-party scanners make requests on your behalf, so confirm the owner is fine with that.
## Step 3: Propose fixes
- Ship headers through the place the stack already uses (framework config, `_headers`, middleware). One review PR.
- CSP in report-only first; switch to enforce only after a clean period and owner approval.
- Remove exposed files from the deploy output and add them to ignore rules; rotation of any leaked secret is the owner's job.
## Validation
- Re-request headers on the preview URL in a real browser; confirm no console CSP violations on key templates (home, article, form, checkout/login).
- Forms still work end to end in a test mode that sends nothing real (or with the owner's test inbox).
- Build and tests pass.
## Report format
Table: finding, severity (high/medium/low/info), evidence (URL, header, response code; never secret values), fix, where it is applied, needs-approval yes/no. Then: what was not tested and why, and the PR link marked "not merged".
## References
- OWASP Secure Headers Project: https://owasp.org/www-project-secure-headers/
- OWASP ASVS 5.0: https://asvs.owasp.org/
- OWASP Cheat Sheets (CSP, CSRF, Session Management, HTTP Headers): https://cheatsheetseries.owasp.org/
- MDN CSP: https://developer.mozilla.org/en-US/docs/Web/HTTP/Guides/CSP
- MDN Strict-Transport-Security and Set-Cookie: https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Headers/Strict-Transport-Security , https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/Headers/Set-Cookie
- Subresource Integrity: https://developer.mozilla.org/en-US/docs/Web/Security/Subresource_Integrity
- RFC 9116 security.txt: https://www.rfc-editor.org/rfc/rfc9116
- Mozilla HTTP Observatory: https://developer.mozilla.org/en-US/observatory
- Cloudflare Turnstile server-side validation: https://developers.cloudflare.com/turnstile/get-started/server-side-validation/
Check before you send
- It gave a plain-language report before suggesting or making any change
- Each finding names the exact page or file and why it matters
- It said what it could not check or verify
- Nothing was changed, published, or deleted without your OK
- Fixes are small and keep the existing design and content
Privacy note: what not to paste
Never paste passwords, API keys, admin URLs with tokens, or .env contents into an AI chat. If the check finds an exposed secret, note where it is and rotate it; don’t copy it anywhere. Only check sites you or your employer control.