Folders and files
148 offensive security skills · 1,569 payload files · 4 agent integrations · Recon-first RED TEAM pipeline
Red-team offensive security engine — deep recon first, signal-gated exploitation, aggressive chaining to critical impact.
claude-hunterKit is an open-source red-team offensive security engine — 148 structured skill files, 1,569 payloads, and a recon-first conditional launch pipeline with 31 escalation paths and 8 stop conditions. Every finding is a primitive to chain into critical impact.
Phase 1: Deep recon (11 skills) → Phase 2: Decision matrix matches 40+ signals → Phase 3: Targeted exploit → Phase 4: Chain & escalate → Phase 5: Report
Every finding is a PRIMITIVE, not a report entry.
- XSS is not "reflected XSS confirmed" — it's a session hijack vector
- SQLi is not "error-based injection" — it's a database exfil pipeline
- SSRF is not "blind SSRF detected" — it's a cloud metadata extraction tunnel
Chain LOW/MEDIUM primitives into CRITICAL impact. A finding without a chain path is an incomplete finding.
Target → PHASE 0: Categorize (web/api/auth/ai/network/cloud)
↓
PHASE 1a: Network/Stack Recon (recon-01,02,03)
PHASE 1b: Surface Mapping (recon-04,05,06)
PHASE 1c: Defense Analysis (recon-07,08,09,10,11)
↓
PHASE 2: Decision Matrix (40+ signal rules → select agents)
↓
PHASE 3: Targeted Exploit (confirm → escalate → exfil)
↓
PHASE 4: Chain & Escalate (31 escalation paths)
↓
PHASE 5: Report (human-voice, impact-first)
| Phase | What Happens | Skills Used |
|---|---|---|
| 0. Categorize | Classify target (web, API, AI, network, cloud, auth) | domain orchestrator |
| 1a. Network/Stack | Subdomains, IPs, WAF, tech stack, CDN, server fingerprint | recon-01 recon-02 recon-03 |
| 1b. Surface Mapping | Crawl, JS endpoints, OpenAPI specs, hidden routes, forms | recon-04 recon-05 recon-06 |
| 1c. Defense Analysis | WAF type, auth schemes, security headers, CORS, shadow APIs | recon-07 recon-08 recon-09 recon-10 recon-11 |
| 2. Decision Matrix | Map 40+ signals → launch matched agents only | _hunter/recon-decision-matrix.yaml |
| 3. Exploit | confirm/ → parameters/ → payloads/ → post-exploit extraction | matched vuln skills |
| 4. Chain | 31 escalation paths: SQLi→RCE, SSRF→Cloud, XSS→ATO, JWT→Admin | escalation_paths in matrix |
| 5. Report | ✅ Confirmed / 🔍 Signal / ❌ Not found — impact-first | reporting skill |
Cardinal rule: Complete ALL of Phase 1 before launching ANY vulnerability agent. Launch agents ONLY for vuln classes where recon found matching signals.
Skills reference wordlists via repo-root-relative paths (wordlists/web/...). The only working layout is skills + wordlists as siblings — either via symlink or the installer.
Clone once, symlink into each CLI's skills directory. The repo stays in place; updates are git pull.
git clone https://github.com/vigilantshield/Claude-HunterKit.git
cd Claude-HunterKit
# Claude Code — symlink skills + wordlists + chrome-devtools MCP
ln -s "$PWD/skills" ~/.claude/skills/claude-hunterkit
ln -s "$PWD/wordlists" ~/.claude/skills/wordlists
ln -s "$PWD/bugbounty" ~/.claude/skills/bugbounty
claude mcp add --scope user chrome-devtools -- npx -y chrome-devtools-mcp@latest
# Gemini CLI — link repo as extension + MCP
gemini extensions link "$PWD" --consent
gemini mcp add -s user chrome-devtools npx -y chrome-devtools-mcp@latest
# Command Code (cmd) — install skill set from GitHub + link wordlists + MCP
cmd skills add vigilantshield/Claude-HunterKit -g
ln -s "$PWD/wordlists" ~/.commandcode/skills/wordlists
cmd mcp add chrome-devtools -- npx -y chrome-devtools-mcp@latest
# Codex CLI — chrome-devtools MCP only (skills referenced from repo)
codex mcp add chrome-devtools -- npx -y chrome-devtools-mcp@latestgit clone https://github.com/vigilantshield/Claude-HunterKit.git
cd Claude-HunterKit
bash install.shDoes the same as above but auto-detects which CLIs are on your PATH and runs the right commands for each. Idempotent — re-running is safe.
| Flag | What It Installs |
|---|---|
| (none) | All detected CLIs: Claude Code → symlinks, cmd → wordlist link, gemini → extension link, codex → MCP |
| --plugin claude | Claude Code only |
| --plugin cmd | Command Code only |
| --plugin gemini | Gemini CLI only |
| --plugin codex | Codex CLI only |
| --mcp | Chrome DevTools MCP only (all CLIs) |
| --domain <name> --target <dir> | Copy one domain's skills + wordlists as siblings to <dir> |
| --dry-run | Preview what would happen (no changes) |
| --list | Show available domains and plugins |
# Bug bounty? Start here:
bugbounty/bugbounty-master/SKILL.md
# Hunting SQLi?
skills/web/web-01-sqli/SKILL.md
# Testing an API?
skills/api/api-01-spec-ingestion/SKILL.md
# AI/LLM chatbot?
skills/ai/ai-01-prompt-injection/SKILL.md
# Need to map the target first?
skills/recon/recon-03-tech-fingerprinting/SKILL.md
# Want the RED TEAM workflow?
skills/_hunter/_hunter-orchestrator.yaml| Metric | Value | What It Means |
|---|---|---|
| 148 | Skills | Every vulnerability class, framework, and technique — deduplicated and clean |
| 1,569 | Payload files | Ready-to-fire wordlists — no external tools needed |
| 51 | Web vuln types | SQLi, XSS, SSRF, IDOR, XXE, SSTI, RCE, GraphQL, JWT... |
| 35 | API skills | BOLA, BFLA, gRPC, WebSocket, mTLS, supply chain... |
| 24 | AI/LLM skills | Prompt injection, jailbreaking, RAG poisoning, MCP... |
| 21 | Network skills | Cloud, K8s, CI/CD, Kerberos, LDAP, serverless... |
| 11 | Recon skills | OSINT, tech fingerprint, JS analysis, OpenAPI, crawl, WAF, auth, headers, CORS, API surface |
| 42,000+ | Lines of methodology | Deep analysis — not shallow checklists |
| 31 | Escalation paths | Chain primitives into critical impact (was 7) |
| 8 | Stop conditions | Know when you own the target (ATO, RCE, PII, IAM keys, pivot) |
| 4 | Agent plugins | Claude Code, Codex CLI, Gemini CLI, cmd |
| 1 | MCP Server | Chrome DevTools (browser automation) |
| 40+ | Signal rules | Decision matrix entries mapping recon findings → vuln agents |
| Other Toolkits | claude-hunterKit |
|---|---|
| Fire all payloads at every target | Recon-first — deep recon decides which agents to launch |
| Payload dumps with no context | Full methodology — 6 sections per skill, consistent, reusable |
| Generic scanning tools | Domain-specific — web vs API vs AI vs network all separated |
| Report findings, not impact | RED TEAM MINDSET — every finding is a chain primitive, not a report entry |
| No chain guidance | 31 escalation paths — built-in attack trees (SQLi→RCE, SSRF→Cloud, XSS→ATO, JWT→Admin) |
| No kill discipline | 8 stop conditions — know when you own the target |
| Agent plugin support | 4 integrations — Claude Code, Codex CLI, Gemini CLI, cmd |
┌─────────────────────────────────────┐ │ 🔮 Hacker Mindset │ ← Think like the attacker │ 🔍 Detection │ ← How to find it │ 📦 Wordlist Payloads │ ← Actual payloads, ready to fire │ ⚡ Bypass Techniques │ ← How to evade filters/WAFs │ ⛓️ Chaining & Escalation │ ← Turn low into critical │ 📡 OOB Detection & Blind Confirm │ ← Prove it without output └─────────────────────────────────────┘
claude-hunterKit/ │ ├── .claude/ # Claude Code config │ ├── CLAUDE.md # RED TEAM pipeline instructions │ └── settings.local.json # MCP server enablement │ ├── .mcp.json # MCP server definition (chrome-devtools) │ ├── bugbounty/ │ └── bugbounty-master/ # Full bug bounty pipeline │ ├── skills/ │ ├── _hunter/ # Master orchestrator + routing │ │ ├── _hunter-orchestrator.yaml # ← RED TEAM entry point │ │ ├── routing-table.yaml │ │ └── recon-decision-matrix.yaml # ← 40+ signal rules, 31 escalation paths │ │ │ ├── recon/ # 11 recon skills │ │ ├── recon-01-osint # OSINT, subdomains, emails, leaks │ │ ├── recon-02-osint-methodology # Structured OSINT │ │ ├── recon-03-tech-fingerprinting # Server, framework, WAF │ │ ├── recon-04-js-analysis # JS endpoints, source maps │ │ ├── recon-05-openapi-enum # API spec discovery │ │ ├── recon-06-crawl-deep # Hidden routes, SPA crawling │ │ ├── recon-07-waf-detection # WAF type, block pages, gaps │ │ ├── recon-08-auth-mapping # JWT/OAuth/SAML/Session │ │ ├── recon-09-security-headers # CSP, HSTS, XFO, CORS │ │ ├── recon-10-cors-scan # Origin reflection, credentialed │ │ └── recon-11-api-surface # Shadow APIs, all endpoints │ │ │ ├── web/ # 51 web skills (deduplicated) │ │ ├── web-01-sqli # SQL injection │ │ ├── web-16-xss # Cross-site scripting │ │ ├── web-27-ssrf # SSRF │ │ ├── web-23-idor # IDOR │ │ ├── web-37-rce # Remote code execution │ │ ├── web-32-graphql # GraphQL │ │ ├── web-34-business-logic # Business logic flaws │ │ ├── web-68-nextjs # Next.js specific │ │ ├── web-69-laravel # Laravel specific │ │ └── ... up to 51 │ │ └── _orchestrator/ # Offensive web orchestrator │ │ │ ├── api/ # 35 API skills │ │ ├── api-01-spec-ingestion # OpenAPI/Swagger │ │ ├── api-04-bola # Broken Object Level Auth │ │ ├── api-10-sqli # API SQL injection │ │ ├── api-16-grpc # gRPC security │ │ ├── api-29-jwt # JWT attacks │ │ ├── api-32-graphql # GraphQL │ │ └── ... up to 35 │ │ └── _orchestrator/ # Offensive API orchestrator │ │ │ ├── ai/ # 24 AI/LLM security skills │ │ ├── ai-01-prompt-injection # Direct/indirect injection │ │ ├── ai-04-system-prompt-leakage │ │ ├── ai-05-rag-poisoning │ │ ├── ai-08-excessive-agency │ │ ├── ai-10-mcp-security │ │ └── ... up to 24 │ │ │ ├── network/ # 21 network/infra skills │ ├── cloud/ # 2 cloud security skills │ ├── auth/ # 2 auth skills │ │ └── _orchestrator/ # Offensive auth orchestrator │ ├── devtools/ # Chrome DevTools MCP │ └── reporting/ # Human-voice report writing │ ├── plugin/ # Agent integration plugins │ ├── .claude-plugin/ # Claude Code │ ├── .codex-plugin/ # Codex CLI │ └── .gemini/ # Gemini CLI │ ├── .commandcode/ # Command Code (cmd) plugin │ ├── wordlists/ # 1,569 payload files │ ├── web/ (583 files) │ ├── api/ (410 files) │ ├── ai/ (387 files) │ ├── network/ (71 files) │ ├── auth/ (100 files) │ ├── recon/ (18 files) │ └── shared/ (2 files — index + invocation matrix) │ ├── gemini-extension.json # Gemini CLI extension manifest ├── install.sh # Universal installer └── README.md
| Skill | Why It Pays |
|---|---|
| SSRF → Cloud Metadata → IAM Keys → Full Account | Critical, fast, reliable |
| Business Logic → Coupon Race → Negative Total → Free Money | Direct financial impact |
| Prompt Injection → System Prompt → API Keys → Full Access | New, unsaturated |
| IDOR/BOLA → User A reads User B's data → PII Breach | Most common API bug |
| JWT → alg:none → Admin Access | One line of JSON = admin |
| CORS Misconfig → Data Exfiltration via Origin Reflection | PII theft via JS |
| WAF Gap → Unprotected endpoint → Direct injection | Bypass all blocks |
| Shadow API → Undocumented endpoint → No auth → Data access | 10x more endpoints |
| Auth Mapping → Public endpoint with sensitive data → IDOR | Every app has these |
| Security Headers → Missing CSP → XSS is game on | Free findings |
Only one MCP server is wired — chrome-devtools (a real, published npm package). Skill routing is handled by the _hunter orchestrator and the CLAUDE.md pipeline, so no separate router server is needed.
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": ["-y", "chrome-devtools-mcp@latest"],
"description": "Browser automation for XSS, SSRF, OAuth verification"
}
}
}| Role | Why You Need This |
|---|---|
| Bug bounty hunters | Full pipeline: recon → signal match → exploit → chain → report |
| Red teamers | RED TEAM MINDSET, 31 escalation paths, 8 stop conditions |
| Pentesters | 51 web vulns + 35 API skills + comprehensive kill chains |
| AI security researchers | 24 LLM-specific skills (most in any open-source repo) |
| SOC / Blue team | Know your enemy — offensive depth for defensive understanding |
0. bugbounty/bugbounty-master/ → Define scope, understand target 1. skills/recon/ → Deep recon (11 skills, 3 phases) 2. _hunter/recon-decision-matrix.yaml → Match signals → select agents 3. Each matched agent: → confirm → escalate → post-exploit 4. Chain findings: → SSRF→Cloud, XSS→ATO, SQLi→RCE, JWT→Admin 5. Check stop conditions: → Hit one? Deliver the chain. 6. skills/reporting/ → Write human-voice report → Get paid ```--- ## 🧪 Verify Installation ### Claude Code ```bash # Skills and wordlists in place? ls ~/.claude/skills/claude-hunterkit/web && echo "✓ skills" ls ~/.claude/skills/wordlists/web/web-03-sqli && echo "✓ wordlists resolve" # MCP server connected? claude mcp list | grep chrome-devtools # → chrome-devtools: npx -y chrome-devtools-mcp@latest - ✔ Connected # Count installed ls ~/.claude/skills/claude-hunterkit/ | wc -l # → should show domain dirs find ~/.claude/skills/claude-hunterkit -name SKILL.md | wc -l # → 148
# Wordlists resolve from installed skills?
ls ~/.commandcode/skills/wordlists/web/web-03-sqli && echo "✓ wordlists resolve"
# Skills installed?
cmd skills list | head -5
# MCP server enabled?
cmd mcp list | grep chrome-devtools# Extension linked?
gemini extensions list | grep claude-hunterkit
# MCP server connected?
gemini mcp list | grep chrome-devtools# MCP server enabled?
codex mcp list | grep chrome-devtoolsbash install.sh --list # 10 domains, plugins, counts
bash install.sh --dry-run # preview (no changes)
find skills/ -name SKILL.md ! -path "*/_*" | wc -l # → 148
find wordlists/ -name "*.txt" | wc -l # → 1,569
find skills/web -name SKILL.md | wc -l # → 51
find skills/recon -name SKILL.md | wc -l # → 11
ls skills/ai/ | wc -l # → 24
ls skills/api/ | wc -l # → 35
# Red team pipeline: verify decision matrix
grep -c "from:" skills/_hunter/recon-decision-matrix.yaml # → 31 escalation paths
grep -c "signals:" skills/_hunter/recon-decision-matrix.yaml # → 40+ signal rulesApache 2.0 — free for commercial and personal use.
⭐ Star this repo if you use it — it helps others find it.
148 skills · 1,569 payloads · 4 agent plugins · 31 escalation paths · RED TEAM MODE