pentest-ai-agents

pentest-ai-agents

50 Claude Code subagents for penetration testing.

Quick Start | Cheatsheet | Coverage | Agents | Examples

---

Table of Contents

• What's New in v3.3
• What's New in v3.2
• Agent Map
• What's New in v3.1
• Quick Start
• Cheatsheet
• Coverage
• Agents
• Tier 1 vs Tier 2
• Examples
• Running Tools in a Container
• Findings Database
• Token Optimization
• Local Models
• Documentation
• MCP Server
• Prerequisites
• Legal
• License

---

pentest-ai-agents is a collection of 50 Claude Code subagents that turn Claude into an offensive security research assistant. Each agent carries deep domain knowledge in a specific area: recon, web, Active Directory, cloud, mobile, wireless, social engineering, payload crafting, reverse engineering, exploit chaining, detection engineering, forensics, and more.

Install the agent files. Open Claude Code. Describe your task. Claude routes to the right specialist automatically.

No servers, no Python deps, no setup beyond copying files.

What's New in v3.3

• Installable as a Claude Code plugin. Two lines — /plugin marketplace add 0xSteph/pentest-ai-agents then /plugin install pentest-ai-agents@pentest-ai-agents. The install.sh curl path still works unchanged.
• 15 new agents (35 → 50): ai-recon (AI attack-surface mapping), code-auditor, crypto-analyzer, password-auditor, database-attacker, network-attacker, traffic-analyzer, compliance-mapper, risk-scorer, plus the post-exploitation set — evasion-specialist, persistence-planner, data-exfiltrator, scada-attacker, iot-pentester, lateral-movement. Every offensive agent pairs its techniques with the detection they exercise.
• Hardened CI validator. A SHA-pinned, least-privilege workflow validates each agent's frontmatter, requires the scope-guard block on every Bash-capable (Tier 2) agent, checks the plugin manifests, and smoke-tests the installer.
• Scope-guard gap closed. cicd-redteam is Bash-capable but was missing the mandatory scope-enforcement block — now fixed (the new CI check would have caught it).
• Installer fixes. curl | bash no longer crashes under set -u, the one-liner clone URL is corrected, slash commands now install alongside the agents, and --uninstall removes everything cleanly.
• Minimal offline Docker bundle with a digest-pinned base and non-root user — packaging only, no tooling baked in.

What's New in v3.2

• 4 new agents: c2-operator (Sliver/Mythic/Havoc/Cobalt Strike profile tuning, beacon hygiene, redirector design), container-breakout (Docker/K8s escape, runc/cri-o CVEs, kubelet exploitation, RBAC abuse), opsec-anonymizer (operator-side identity hygiene, source IP design, burner infrastructure, fingerprint hygiene), llm-redteam (OWASP LLM Top 10 testing, prompt injection, RAG poisoning, MCP server abuse, agent tool abuse).
• Tightened scope guard: explicit hard-refusal list in _scope-guard.md covers DoS, mass scanning, unattended worms, false-flag operations, safety-of-life systems.
• Findings DB v2: vulns.tool_used column for filtering findings by the tool that produced them; new indexes on cve and tool_used. Existing engagements migrate forward via db/migrate.sh.
• Agent map diagram: visual flow from recon to closure mapped to agent names (see below).

Agent Map

flowchart LR
    classDef plan fill:#1a2a4a,stroke:#5a7ab8,color:#eaf0ff
    classDef recon fill:#1a3a2a,stroke:#5ab87a,color:#eaffea
    classDef exploit fill:#3a1a1a,stroke:#b85a5a,color:#ffeaea
    classDef post fill:#3a2a1a,stroke:#b8895a,color:#fff0ea
    classDef defense fill:#1a3a3a,stroke:#5ab8b8,color:#eaffff
    classDef report fill:#2a1a3a,stroke:#895ab8,color:#f0eaff

    EP[engagement-planner]:::plan
    OA[opsec-anonymizer]:::plan
    TM[threat-modeler]:::plan

    OS[osint-collector]:::recon
    RA[recon-advisor]:::recon
    VS[vuln-scanner]:::recon

    WH[web-hunter]:::exploit
    AS[api-security]:::exploit
    BL[bizlogic-hunter]:::exploit
    BB[bug-bounty]:::exploit
    AD[ad-attacker]:::exploit
    CS[cloud-security]:::exploit
    MP[mobile-pentester]:::exploit
    WP[wireless-pentester]:::exploit
    LR[llm-redteam]:::exploit
    SE[social-engineer]:::exploit
    PO[phishing-operator]:::exploit
    CT[ctf-solver]:::exploit
    CR[credential-tester]:::exploit
    PV[poc-validator]:::exploit
    EG[exploit-guide]:::exploit
    EC[exploit-chainer]:::exploit
    AP[attack-planner]:::exploit
    PC[payload-crafter]:::exploit
    RE[reverse-engineer]:::exploit

    PE[privesc-advisor]:::post
    CB[container-breakout]:::post
    C2[c2-operator]:::post
    CI[cicd-redteam]:::post
    SO[swarm-orchestrator]:::post

    DE[detection-engineer]:::defense
    FA[forensics-analyst]:::defense
    MA[malware-analyst]:::defense
    SA[stig-analyst]:::defense

    RG[report-generator]:::report

    EP --> OA --> OS
    EP --> TM
    OS --> RA --> VS
    VS --> WH & AS & BL & BB & AD & CS & MP & WP & LR
    SE --> PO
    BB --> WH
    PO --> PC --> C2
    AD --> CR
    AD --> PE
    CS --> CB
    CB --> PE
    WH --> PV
    AS --> PV
    PV --> EC --> AP
    EC --> EG
    PC --> RE
    RE --> MA
    AP --> SO
    C2 --> DE
    SO --> RG
    DE --> FA
    MA --> RG
    SA --> RG
    CT -.solo.-> RG
    CI -.pipeline.-> SO

Tier 1 (advisory) agents are routable from any task. Tier 2 (execution-capable) agents require a declared scope and live in the offensive operations cluster.

What's New in v3.1

• 3 new agents: payload-crafter (msfvenom, Donut, custom loaders), reverse-engineer (Ghidra, JadX, Radare2, Binwalk), phishing-operator (Evilginx, GoPhish, dnstwist)
• Slash commands: /recommend "freeform task" routes you to the right agent + concrete commands. /agents-for <tag> filters the catalog by domain.
• db/doctor.sh: audits which underlying CLI tools are installed on your box, grouped by agent. Shows ✔ and ✘ per tool with install hints.
• install.sh --tools: opt-in installer that pulls in the underlying tools via apt/brew/pacman + pipx/go/cargo.
• Extended agents: Commix added to web-hunter, RouterSploit added to vuln-scanner, targeted wordlist generation (cupp, CeWL, Mentalist, Crunch, hashid, haiti) added to credential-tester, full steganography toolkit added to ctf-solver.

Quick Start

One command:

curl -fsSL https://raw.githubusercontent.com/0xSteph/pentest-ai-agents/main/install.sh | bash

That's it. The script clones the repo to a temp dir, copies the agents to ~/.claude/agents/, and exits. Idempotent: safe to re-run for updates.

Or install as a Claude Code plugin (no clone; updates through the marketplace):

/plugin marketplace add 0xSteph/pentest-ai-agents
/plugin install pentest-ai-agents@pentest-ai-agents

This registers all 52 agents and the slash commands through Claude Code's plugin system. Pick the plugin or the installer — you don't need both.

Then open Claude Code:

"Plan an internal network pentest for a 500-endpoint AD environment with a 2-week window."

Claude routes to the engagement planner agent and produces a phased plan with MITRE ATT&CK mappings.

Prefer to clone first?

git clone https://github.com/0xSteph/pentest-ai-agents.git
cd pentest-ai-agents && ./install.sh --global

Other install options:

./install.sh --project     # Install for current project only
./install.sh --global --lite  # Use Haiku for advisory agents (lower cost)
./install.sh --tools       # Install underlying CLI tools (nmap, nuclei, ffuf, etc.)
./install.sh --help        # All options

See INSTALL.md for step-by-step instructions, including first-time Claude Code setup.

---

Cheatsheet

Quick interactions once installed:

Command	What It Does
/recommend "phish a small SaaS team's IT department"	Picks the right agent and gives concrete next commands
/agents-for web	Lists every agent relevant to web testing (web-hunter, api-security, bug-bounty, bizlogic-hunter)
/agents-for cloud	Cloud-specific agents (cloud-security, cicd-redteam)
db/doctor.sh	Audits which underlying CLI tools you have, grouped by agent. Shows ✔/✘ and install hints.
db/doctor.sh --agent ad-attacker	Audit just the AD tooling stack
db/doctor.sh --json	Machine-readable output for piping into a script
install.sh --tools	Install the underlying tools via your package manager + pipx/go/cargo
findings.sh init <id>	Start a new engagement (persistent SQLite findings DB)
findings.sh stats	Engagement progress
findings.sh export	Full JSON export
bash handoff.sh	Markdown handoff report between sessions

In Claude Code, just describing your task routes automatically:

"Plan an internal pentest for a 500-endpoint AD environment, 2-week window."
"I have a domain user, where do I look first in BloodHound?"
"Convert this SharpHound EXE into shellcode for an EDR test, with detection content."
"Reverse this firmware image and tell me what the cryptographic protocol looks like."
"Run a phishing simulation against acme-corp.com, set up GoPhish + Evilginx infrastructure."

---

Coverage

What the agents drive. Categories map to the same surface real adversaries operate across:

Category	Agents	Underlying Tools (installable via install.sh --tools)
Recon and OSINT	recon-advisor, osint-collector	nmap, masscan, rustscan, dig, whois, subfinder, amass, httpx, theHarvester, sherlock, holehe, maigret
Vulnerability scanning	vuln-scanner	nuclei, nikto, RouterSploit, nmap NSE, OpenVAS/Nessus parsing
Web app testing	web-hunter, api-security, bug-bounty, bizlogic-hunter	ffuf, gobuster, feroxbuster, sqlmap, dalfox, Commix, dirsearch, whatweb
Active Directory	ad-attacker, credential-tester	BloodHound, Impacket, NetExec/CrackMapExec, Certipy, kerbrute, Responder, ldapsearch
Credentials and cracking	credential-tester, password-auditor	Hydra, Hashcat, John, Medusa, cupp, CeWL, NIST 800-63B policy audit, HIBP k-anonymity
Cloud	cloud-security, cicd-redteam	aws/az/gcloud CLIs, Trivy, Prowler, ScoutSuite, Pacu
Containers and K8s breakout	container-breakout	kubectl, kube-hunter, peirates, CDK, Falco rule pairing
C2 operations	c2-operator	Sliver, Mythic, Havoc, Cobalt Strike, malleable profiles, redirector design
AI / LLM red teaming	ai-recon, llm-redteam	A2A/MCP/RAG surface discovery, model fingerprinting, Garak, PyRIT, Promptfoo, OWASP LLM Top 10, MITRE ATLAS
Operator opsec	opsec-anonymizer	source IP design, burner identity, JA3/fingerprint hygiene, burn checklists
Mobile	mobile-pentester, reverse-engineer	Frida, Objection, jadx, apktool, MobSF, adb