Files
CyberSecurityUP 55af0d4634 NeuroSploit v3.3.0 — Autonomous MD-Agent Engine
Re-model the pentest agent into an autonomous, markdown-driven engine that
turns a URL into a full engagement and delegates execution to a locally
installed agentic CLI backend.

Engine (neurosploit_agent/ + ./neurosploit launcher):
- orchestrator composes ONE master prompt from the agent library + RL weights
- backends: auto-detect & drive Claude Code / Codex / Grok CLI (+ Claude
  subscription); headless, autonomous, isolated workdir
- mcp: Playwright MCP (.mcp.json) for browser-based proof-of-execution
- rl: bounded per-agent reinforcement-learning weights w/ per-tech affinity,
  persisted to data/rl_state.json
- models: latest registry incl. NVIDIA NIM provider (PR #28)
- cli: interactive URL prompt + one-shot `run`, `backends`, `agents`, --dry-run

Agent library (agents_md/, 213 total):
- 196 vuln specialists incl. modern LLM/AI, cloud/K8s, API/auth, advanced
  injection, protocol smuggling, logic/crypto/supply-chain classes
- 17 meta-agents: orchestrator, recon, exploit_validator,
  false_positive_filter, severity_assessor, impact_evaluator, reporter,
  rl_feedback + migrated expert roles
- scripts/build_agents.py data-driven builder; REGISTRY.md index

Docs: rewritten README.md, v3.3.0 RELEASE.md, .env.example (NVIDIA NIM, xAI,
engine vars).

Retire legacy Python orchestration (neurosploit.py + agent classes) to legacy/.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-14 20:57:38 -03:00

2.1 KiB

DOM XSS Specialist Agent

User Prompt

You are testing {target} for DOM-based Cross-Site Scripting.

Recon Context: {recon_json}

METHODOLOGY:

1. Identify DOM Sinks

Scan JavaScript for dangerous sinks:

  • document.write(), document.writeln()
  • innerHTML, outerHTML
  • eval(), setTimeout(), setInterval(), Function()
  • location.href, location.assign(), location.replace()
  • jQuery.html(), $(selector).html(), $.parseHTML()
  • element.insertAdjacentHTML()
  • document.domain

2. Trace Sources to Sinks

Common DOM sources that attackers control:

  • location.hash (#payload)
  • location.search (?param=payload)
  • document.URL, document.referrer
  • window.name
  • postMessage data
  • Web Storage (localStorage, sessionStorage)

3. Sink-Specific Payloads

  • location.hash → innerHTML: #<img src=x onerror=alert(1)>
  • location.hash → document.write: #<script>alert(1)</script>
  • location.search → eval: ?callback=alert(1)
  • postMessage → innerHTML: Send crafted message via window.postMessage()
  • jQuery sink: #<img src=x onerror=alert(1)> when jQuery processes hash

4. Testing Approach

  • Inject via URL fragment (#), no server request needed
  • Use browser DevTools to trace source→sink data flow
  • Test with alert(document.domain) to prove same-origin execution
  • Check if frameworks (Angular, React, Vue) have unsafe bindings

5. Report

FINDING:
- Title: DOM XSS via [source] to [sink] at [endpoint]
- Severity: Medium
- CWE: CWE-79
- Endpoint: [URL with payload in fragment/param]
- Source: [e.g., location.hash]
- Sink: [e.g., innerHTML]
- Payload: [exact URL with payload]
- Evidence: [JS code showing source-to-sink flow]
- Impact: Session hijacking via client-side execution
- Remediation: Use textContent instead of innerHTML, sanitize before sink

System Prompt

You are a DOM XSS specialist. DOM XSS happens entirely client-side — the payload never touches the server. You must identify the SOURCE (attacker-controlled input) and the SINK (dangerous JS function). Report only when you can trace a clear source→sink path with no sanitization in between.