Bug-Bounty-Agents

A curated arsenal of specialized AI agent prompts for bug bounty hunting, penetration testing, and offensive security workflows.

Drop-in personas for Claude Code, Copilot Chat, Cursor, and any agent-capable LLM - no frameworks, no dependencies, just disciplined prompts.

43 agents  ·  6 engagement phases  ·  4 supported clients  ·  0 dependencies

Quick Start  ·  Catalog  ·  Setup  ·  Workflows  ·  Examples  ·  Contributing  ·  Disclaimer

---

Overview

Each .md file in this repository defines a focused, production-ready agent persona - recon, web hunting, exploit chaining, reporting, and more - that you can drop into Claude Code, GitHub Copilot Chat, Cursor, or any agent-capable LLM client.

No frameworks. No dependencies. Just disciplined prompts that turn a generic LLM into a specialist, with strict scope enforcement built in.

These are prompts, not scanners. They make an LLM act like a specialist; they do not bring their own tooling. You still drive the engagement.

---

Table of Contents

1. Quick Start
2. Agent Catalog
3. Prerequisites
4. Per-Tool Setup
   • One-line installer
   • Claude Code
   • GitHub Copilot Chat
   • Cursor
   • ChatGPT / Gemini / Generic
5. Using an Agent
6. Workflows
7. Examples
8. Burp Suite MCP Integration
9. Updating
10. Project Files
11. Contributing
12. Security
13. Disclaimer

---

Quick Start

git clone https://github.com/matty69v/Bug-Bounty-Agents.git
cd Bug-Bounty-Agents
./install.sh                     # auto-detects your client(s)

Or pick a specific target:

./install.sh --target claude         # Claude Code (global)
./install.sh --target claude-local   # Claude Code (this project)
./install.sh --target copilot        # Copilot Chat (VS Code)
./install.sh --target cursor         # Cursor (this project)
./install.sh --target all            # everything detected
./install.sh --dry-run --target claude
./install.sh --uninstall --target claude

---

Agent Catalog

Agents are grouped by phase of an offensive engagement. The full machine-readable index lives in AGENTS.md.

Reconnaissance & Intelligence

Agent	Purpose
recon-advisor	Surface enumeration and asset discovery
osint-collector	Open-source intelligence gathering
subdomain-takeover	Dangling DNS and subdomain takeover validation
threat-modeler	STRIDE / attack-surface modeling
engagement-planner	Scope, rules of engagement, test plans
attack-planner	Multi-stage attack path planning

Web, API & Application

Agent	Purpose
web-hunter	Web application vulnerability hunting
api-security	REST and GraphQL API testing
graphql-hunter	Schema introspection, authz, complexity attacks
bizlogic-hunter	Business logic flaws and abuse cases
ssrf-hunter	SSRF discovery, filter bypass, cloud-metadata abuse
jwt-cracker	JWT / OIDC token attacks (alg confusion, kid/jku, weak HMAC)
vuln-scanner	Automated scanning orchestration and triage

Infrastructure, Cloud & Network

Agent	Purpose
cloud-security	AWS / GCP / Azure misconfiguration hunting
container-escape	Docker / Kubernetes pod-to-node-to-cluster breakout
cicd-redteam	CI/CD pipeline and supply-chain attacks
ad-attacker	Active Directory enumeration and abuse
wireless-pentester	Wi-Fi, Bluetooth, and RF assessments
mobile-pentester	iOS / Android application testing
hardware-hacker	Embedded, JTAG, firmware extraction

Exploitation & Post-Ex

Agent	Purpose
exploit-chainer	Combine findings into impactful chains
exploit-guide	Step-by-step exploitation reference
payload-crafter	Custom payload generation and tuning
binary-exploit	Memory corruption, ROP, pwn
crypto-analyst	Crypto primitive and protocol analysis
credential-tester	Password spraying, stuffing, brute force
privesc-advisor	Linux / Windows privilege escalation paths
poc-validator	Verify, stabilize, and minimize PoCs
red-team-operator	C2, OPSEC, long-haul operations

Specialized & Adversarial

Agent	Purpose
llm-redteam	Prompt injection, tool abuse, RAG poisoning, agent loops
phishing-operator	Phishing infrastructure and campaign design
social-engineer	Pretexting, vishing, human-layer attacks
malware-analyst	Static and dynamic malware analysis
reverse-engineer	Binary RE, decompilation, patching
forensics-analyst	DFIR, artifact analysis, timeline building
ctf-solver	CTF challenge solver across categories

Defense, Reporting & Orchestration

Agent	Purpose
detection-engineer	Detection and response engineering
purple-team	Detection-as-you-attack collaboration
stig-analyst	STIG / CIS / compliance hardening review
report-generator	Triage-ready bug bounty reports
bug-bounty	General-purpose bounty assistant
swarm-orchestrator	Coordinate multiple agents in parallel
_scope-guard	Hard scope enforcement layered on any agent

---

Prerequisites

• git and bash installed on your machine
• An LLM client that supports custom system prompts or instruction files:
  • Claude Code
  • GitHub Copilot Chat (VS Code)
  • Cursor
  • ChatGPT (Custom GPTs / Projects), Gemini, or any chat UI accepting a system prompt

---

Per-Tool Setup

One-line installer

./install.sh           # interactive - detects what you have
./install.sh --help    # see all options

The installer auto-detects claude, code, and cursor on your PATH, copies agents to the correct directory for each, and renames files appropriately (e.g. .chatmode.md for Copilot). Use --dry-run to preview, --uninstall to remove.

Claude Code reads agent definitions from ~/.claude/agents/ (global) or .claude/agents/ (per-project).

# Global
mkdir -p ~/.claude/agents && cp *.md ~/.claude/agents/

# Per-project
mkdir -p .claude/agents && cp /path/to/Bug-Bounty-Agents/*.md .claude/agents/

/agents
> use the web-hunter agent to audit https://target.example.com

Copilot Chat supports custom chat modes via .chatmode.md files.

# macOS
PROMPTS_DIR="$HOME/Library/Application Support/Code/User/prompts"
# Linux:   PROMPTS_DIR="$HOME/.config/Code/User/prompts"
# Windows: %APPDATA%\Code\User\prompts

mkdir -p "$PROMPTS_DIR"
for f in *.md; do
  cp "$f" "$PROMPTS_DIR/$(basename "$f" .md).chatmode.md"
done

Reload VS Code, then select the mode from the Copilot Chat dropdown.

cd /your/project
mkdir -p .cursor/rules
cp /path/to/Bug-Bounty-Agents/*.md .cursor/rules/

Each file becomes a selectable rule in Cursor's chat panel.

Open the agent file, copy its full contents, and paste into:

• ChatGPT - Custom GPT → Instructions, or Project → Instructions
• Gemini - Gem instructions
• Open WebUI / LM Studio - System prompt field
• API clients - system role message

---

Using an Agent

Once installed, give the agent a concrete target and scope:

Target: https://staging.acme.example.com
Scope:  *.acme.example.com (in scope), *.thirdparty.example.com (out)
Goal:   Find auth bypass and IDOR on /api/v2/users endpoints.

Well-behaved agents will:

• Ask clarifying questions before acting
• Stay strictly within scope
• Produce reproducible PoCs
• Output triage-ready findings with severity and impact

---

Workflows

Use swarm-orchestrator or attack-planner to coordinate a full engagement:

flowchart LR
    A[recon-advisor]:::phase --> B[web-hunter<br/>api-security]:::phase
    B --> C[exploit-chainer]:::phase
    C --> D[poc-validator]:::phase
    D --> E[report-generator]:::phase

    A -.- A1([enumerate attack surface]):::note
    B -.- B1([find vulnerabilities]):::note
    C -.- C1([escalate impact]):::note
    D -.- D1([confirm &amp; stabilize]):::note
    E -.- E1([write the submission]):::note

    classDef phase fill:#0d1117,stroke:#30363d,color:#e6edf3,stroke-width:1px;
    classDef note  fill:#00000000,stroke:#00000000,color:#8b949e;

Layer _scope-guard on top of any agent to enforce hard scope boundaries during long-running sessions. For purple-team work, run red-team-operator and purple-team side by side.

---

Examples

End-to-end engagement walkthroughs (sanitized) live in examples/:

• web-bug-bounty.md - recon → web-hunter → bizlogic → chain → validate → report, ending in a Critical-tier HackerOne submission.

---

Burp Suite MCP Integration

PortSwigger's MCP Server lets your LLM client drive Burp Suite directly - issue requests through the proxy, query Repeater/Intruder, read site maps, and pivot off live traffic while an agent in this repo provides the methodology.

Pairing tip: load web-hunter, api-security, ssrf-hunter, or bizlogic-hunter alongside the Burp MCP so the agent can both think like a specialist and act through Burp.

Prerequisites

• Burp Suite (Community or Professional) installed and running
• Java available on PATH (java --version)
• jar available on PATH (jar --version) - required to build
• An MCP-capable client (Claude Desktop, Claude Code, Cursor, etc.)

Build the extension

git clone https://github.com/PortSwigger/mcp-server.git
cd mcp-server
./gradlew embedProxyJar
# output: build/libs/burp-mcp-all.jar

Load into Burp Suite

1. Launch Burp Suite.
2. Go to Extensions → Add.
3. Set Extension Type to Java.
4. Select build/libs/burp-mcp-all.jar and click Next.
5. Open the new MCP tab and tick Enabled.
   • Optional: enable tools that can edit your config if you trust the client.
   • Default listener: http://127.0.0.1:9876.

Wire up your MCP client

Claude Desktop (auto): in the Burp MCP tab, click the installer button - it writes the config for you. Restart Claude Desktop.

Claude Desktop (manual): edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "burp": {
      "command": "/path/to/burp/jre/bin/java",
      "args": [
        "-jar",
        "/path/to/mcp-proxy-all.jar",
        "--sse-url",
        "http://127.0.0.1:9876"
      ]
    }
  }
}

Use the Burp MCP tab's installer to extract mcp-proxy-all.jar if you don't already have it.

SSE-capable clients (Cursor, Claude Code, custom): point them straight at the SSE endpoint - no proxy needed:

http://127.0.0.1:9876/sse

Smoke test

With Burp running, the extension loaded, and your client restarted, ask:

Use the burp MCP to list the last 10 requests in the proxy history,
then pick anything that looks like an authenticated API call.

If the client returns live traffic from your Burp session, you're wired up.

---

Updating

cd ~/path/to/Bug-Bounty-Agents
git pull
./install.sh         # re-runs install with the latest agents

---

Project Files

File	Purpose
README.md	This file
AGENTS.md	Machine-readable index (phase, ATT&CK tactic, risk tier)
CHANGELOG.md	Version history
CONTRIBUTING.md	How to add or update agents
SECURITY.md	How to report prompt-safety issues
LICENSE	MIT
install.sh	Auto-detecting installer
templates/AGENT_TEMPLATE.md	Boilerplate for new agents
examples/	Sanitized engagement walkthroughs
.github/	Issue / PR templates and CI

---

Contributing

PRs and issues are welcome. See CONTRIBUTING.md for the contribution workflow, agent template, and style guide. Use the issue templates for bug reports and new-agent proposals.

---

Security

Found a prompt-safety or supply-chain issue? See SECURITY.md and report privately via GitHub Security Advisories.

---

Disclaimer

These agents are intended for authorized security testing only - bug bounty programs you are enrolled in, systems you own, or environments where you have explicit written permission to test.

Unauthorized testing is illegal in most jurisdictions. You alone are responsible for how you use these prompts.

---

Built for hunters who prefer disciplined prompts over brittle frameworks.

Star on GitHub  ·  Report an issue  ·  Contribute

MIT licensed · Authorized testing only