Dom Security Hardening

DOM Security Hardening

Purpose

XSS attacks kill applications. This skill hardens the DOM attack surface by enforcing a strict Content Security Policy, eliminating unsafe DOM APIs, and stripping execution vectors like inline scripts and styles. Following this skill is MANDATORY for any user-facing web application handling user input.

When to use

• Setting up the initial index.html or document root of a web application
• Refactoring legacy code that relies on direct DOM manipulation
• Auditing a frontend codebase for XSS vulnerabilities
• Processing rich text or markdown input from untrusted users
• Creating forms or chat systems that accept user input

When NOT to use

• CSS framework design (separate concern)
• React component prop validation (use React Component Design)
• Backend input validation (different layer, still required)

Inputs required

• HTML document structure
• JavaScript files with DOM manipulation
• Understanding of Content Security Policy basics
• Sanitizer library decision (DOMPurify, sanitize-html, etc.)

Workflow

1. Enforce CSP: Add strict ` tag in HTML ` (or HTTP headers)
2. Externalize Assets: Extract ALL inline ` and blocks into separate .js and .css` files
3. Remove Inline Events: Replace inline onclick="...", onchange="..." with standard addEventListener bindings
4. Replace Unsafe DOM: Replace innerHTML, outerHTML, dangerouslySetInnerHTML with textContent or innerText
5. Implement Sanitizer: If HTML rendering is REQUIRED, use DOMPurify or equivalent BEFORE insertion
6. Test CSP: Verify policy blocks all unauthorized execution attempts
7. Verify No Bypasses: Use security scanner (OWASP ZAP, Burp) to confirm XSS vectors are closed

Rules

• MUST implement strict CSP: default-src 'self'; script-src 'self'; style-src 'self'; object-src 'none'
• MUST NEVER use inline `` blocks in HTML
• MUST NEVER use inline style="..." attributes
• MUST NEVER use eval(), setTimeout(string), new Function(string)
• MUST NEVER use innerHTML, outerHTML, or dangerouslySetInnerHTML with user input
• MUST use textContent or innerText for all dynamic text insertions
• MUST sanitize user HTML with DOMPurify before any insertion
• MUST NOT use 'unsafe-inline' or 'unsafe-eval' in CSP

Anti-patterns

• innerHTML Assignment: element.innerHTML = userInput (immediate XSS)
• javascript: URIs: href="javascript:void(0)" or href="javascript:alert(1)"
• Unsafe CSP: Content-Security-Policy: default-src *; script-src 'unsafe-inline'
• Event Handler Strings: Creating event handlers from user input or strings
• Trusting User Input: Assuming any user input is safe to insert into DOM
• Missing Sanitizer: Using rich text editor without sanitizing output

Failure conditions

• CSP header/meta tag is missing
• Inline scripts or styles remain in production code
• innerHTML used with user input
• dangerouslySetInnerHTML used without sanitization
• CSP allows 'unsafe-inline' or 'unsafe-eval'
• Sanitizer library is not installed for rich text rendering

Validation checklist

• [ ] CSP header/meta tag present with restrictive policy
• [ ] No `` tags with inline code (all external)
• [ ] No inline style="..." attributes (all CSS classes)
• [ ] No onclick, onchange, oninput inline event handlers
• [ ] All DOM text insertions use textContent or innerText
• [ ] No innerHTML, outerHTML, or dangerouslySetInnerHTML with user input
• [ ] DOMPurify (or equivalent) used for any rich text rendering
• [ ] No eval(), setTimeout(string), new Function(string) calls
• [ ] Security scanner passes (no XSS vulnerabilities detected)
• [ ] CSP blocks inline script execution (verify in browser console)

Output format

• HTML file: Strict CSP meta tag in `, external tags at end of `
• JavaScript: All DOM mutations via safe APIs (textContent, className, setAttribute)
• CSS: Separate .css files, no inline styles anywhere
• Rich Text: HTML sanitized via DOMPurify before insertion
• Validation: Automated XSS scan passes

Security considerations

• Threat Model: Prevent XSS attacks via user input injection, DOM gadgets, third-party scripts
• Mitigations: CSP prevents execution, sanitizer prevents HTML injection, safe APIs prevent eval
• Constraints: CSP may conflict with analytics/ads (allow specific domains only)
• Legacy Code: Some frameworks may require CSP relaxation (document trade-offs)
• Third-party Scripts: Load analytics/ads only from trusted CDNs with subresource integrity (SRI)

Agent execution notes

• Agent MAY: Add CSP header, externalize inline scripts/styles, replace innerHTML with textContent, implement DOMPurify
• Agent MUST NEVER: Use 'unsafe-inline' or 'unsafe-eval' in CSP, ignore sanitization requirements, leave inline event handlers
• Agent MUST ASK: Before adding third-party scripts, before relaxing CSP for legacy code, before using dangerouslySetInnerHTML
• Agent MUST VALIDATE: CSP policy is strict, no inline scripts remain, no unsafe DOM APIs, XSS scanner passes

Example

❌ Anti-pattern (Unsafe XSS vectors, no CSP):

Click me

  // Unsafe: direct user input to DOM
  document.getElementById('user-bio').innerHTML = getUserInput();
  
  // Unsafe: inline style
  document.getElementById('user-bio').setAttribute('style', 'color: red;');
  
  // Unsafe: event handler string
  setTimeout("console.log('vulnerable')", 1000);

✅ Correct pattern (Hardened, safe):

Click me

// app.js

// 1. Safe event binding
document.getElementById('submit-btn').addEventListener('click', (e) => {
  e.preventDefault();
  submitForm();
});

// 2. Safe text insertion (escapes HTML by default)
document.getElementById('user-bio').textContent = getUserInput();

// 3. Safe styling via CSS classes
document.getElementById('user-bio').classList.add('text-red');

// 4. If HTML rendering is absolutely required, sanitize first
import DOMPurify from 'dompurify';
document.getElementById('user-bio').innerHTML = DOMPurify.sanitize(getUserMarkdown());

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: KraitDev
• Source: KraitDev/skiLL.Md
• License: MIT

Install and usage instructions live in the source repository linked above.