Shell WalkerIntermediate21 daysFree

Bug Bounty Hunting with Claude and ChatGPT in 21 Days

Use Claude and ChatGPT as sharp, well-supervised assistants to find, prove and report real vulnerabilities in 21 days of hands-on practice.

Taught by Mr Elite 14 of 21 days live A new day every day

Start Day 1 See the syllabus

0of 21 days done

This is a 21-day course on finding and reporting vulnerabilities the way bounty hunters do, with Claude and ChatGPT working alongside you. I do not treat the assistants as magic. They are fast, confident and sometimes wrong, so you will learn where they save you hours and where they invent things. You will plan, test, verify and write up, and you stay in charge of every decision.

Each day is one lesson of 45 to 75 minutes including practice, and each ends with a mission that produces something I can check: a screenshot, a note, a working command or a short write-up. Everything hands-on runs on your own machine or in the SecurityElites practice labs. Nothing in this course points you at a real company's systems. We cover web bugs, then bugs in AI-powered features, then reporting, and we finish with a capstone that uses all of it.

By the end you can

  • Run a safe, scoped bounty workflow that uses AI for planning, explaining and drafting while you verify everything yourself.
  • Redact sensitive data before pasting it into an assistant and spot when an assistant is guessing.
  • Find and prove common web flaws in labs, including XSS, IDOR, broken authentication, SSRF, injection and logic bugs.
  • Test AI-powered features for prompt injection, data leaks and unsafe handling of model output.
  • Write clear, reproducible reports with honest severity and impact, then handle triage replies professionally.
  • Build a personal playbook of prompts, checklists and notes that you can reuse on real programs within their rules.

Who this is for

  • You are comfortable in a terminal and understand how HTTP, cookies and basic networking work.
  • You have tried a few labs or capture-the-flag challenges and want a repeatable bug bounty method.
  • You already use, or are curious about, Claude and ChatGPT and want to use them safely and effectively for security work.
  • You do not need previous bounty experience or any payouts to begin.

Before you start

  • Comfort with a command line (navigating folders, running commands, reading output).
  • A working idea of how web requests, responses, cookies and status codes behave.
  • A free account with Claude and a free account with ChatGPT.
  • A computer where you can install Burp Suite Community Edition and Python 3.
  • Claude (free account)
  • ChatGPT (free account)
  • Burp Suite Community Edition
  • A modern browser with developer tools
  • curl and jq
  • Python 3
  • Any plain text editor

Syllabus

One day at a time. Finished days get a tick. Days that are not out yet show when they unlock.

Module 3 Server-Side and Logic Bugs

Test SSRF, OAuth, injection into servers, CSRF and logic flaws while using AI to reason about what the app is doing.

  1. 8Module Review: APIs, CORS and GraphQL HuntBy the end of today you can assess an API for CORS and GraphQL weaknesses while using earlier web techniques.Review | Lab: CORS Misconfiguration | Lab: GraphQL

Module 5 Reports, Playbook and Capstone

Turn findings into strong reports and a reusable method, then prove it end to end.

  1. 13Writing Reports That Get Triaged WellBy the end of today you can write a clear, reproducible bounty report and use an assistant to tighten it without adding false claims.Practice | Lab: IDOR
  2. 14Triage, Duplicates and Your Personal PlaybookBy the end of today you can respond to triage feedback calmly and assemble a reusable playbook of prompts and checklists.Review
  3. 15Broken Authentication: JWTs and Password ResetsBy the end of today you can decode a JWT, test a none-algorithm bypass, and assess a password reset token for predictability.Practice | Lab: JWT | Lab: Password Reset | Coming soon
  4. 16Module Review: Injection Basics and a Mixed Web RunBy the end of today you can combine earlier web skills with SQL and NoSQL injection and choose the right technique for an unknown login form.Review | Lab: SQL Injection | Lab: NoSQL Injection | Coming soon
  5. 17SSRF: Making the Server Fetch for YouBy the end of today you can prove a basic SSRF flaw by making an app fetch an internal address in the lab.Practice | Lab: SSRF | Coming soon
  6. 18OAuth Flaws: Redirects and Missing StateBy the end of today you can find redirect_uri and state parameter weaknesses in an OAuth login flow.Practice | Lab: OAuth | Lab: OAuth | Coming soon
  7. 19From Input to Server Execution: Command and Template InjectionBy the end of today you can detect and prove command injection and server-side template injection in lab applications.Practice | Lab: Command Injection | Lab: Server-Side Template Injection | Coming soon
  8. 20Logic Flaws: Races and Cross-Site Request ForgeryBy the end of today you can show a race condition and a CSRF weakness in a lab and explain the business impact.Practice | Lab: Race Condition | Lab: CSRF | Coming soon
  9. 21Capstone: Two Labs, Full Workflow, and What NextBy the end of today you can run the whole workflow from scope to report on two labs and plan your next steps.Capstone | Lab: AI Agent Hijacking | Lab: JWT | Coming soon

The final project

You take two SecurityElites labs from different categories and run the full workflow on each: plan with an assistant, test, verify, and write a report in bounty format. You finish with a personal playbook and a plan for entering real programs within their written rules.

Questions people ask

Do I need paid accounts for Claude or ChatGPT?

No. Free accounts are enough for everything in this course. Paid tiers may give longer conversations, but none of the missions depend on them.

Will I test real companies during the course?

No. Every hands-on step runs on your own machine or in the SecurityElites labs. Real programmes are only something you read about, and when you are ready to test one you must follow its written scope.

Can an AI assistant find bugs for me?

It can speed up reading code, explaining traffic and drafting reports, but it also makes confident mistakes. You will learn to verify every claim with a real request before you rely on it.

Is it a problem if I get stuck on a lab?

Not at all. Ask an assistant for a hint rather than the answer, check what it says against the lab behaviour, and write down what you tried. That note is part of the learning.