Tip

Be with us!

	Join our Discord community to connect with contributors.
	Updates posted by @justsisyphus.

---

Oh My Open Pentest

An AI agent that runs penetration tests. End to end. No babysitting.

Define the scope. Say fullscan. The agent runs recon, enumerates the attack surface, exploits findings, verifies them, and delivers a submission-ready report. You review results, not progress.

Human intervention during an engagement is a failure signal. If the system is designed correctly, the agent completes the cycle — recon through report — without requiring babysitting.

— Manifesto

---

What it does

You:     fullscan https://target.example.com
Agent:   [RECON]   subfinder, httpx, nmap, theHarvester...
         [ENUM]    nuclei, ffuf, nikto, arjun, katana...
         [EXPLOIT] sqlmap, dalfox, netexec, certipy...
         [VERIFY]  every finding confirmed before report
         [REPORT]  CVSS scores, PoCs, reproduction steps
You:     read the report

109 tools. 250 skill playbooks. 10 engagement modes. The agent picks the right tools for the target — you don't configure anything.

---

v2.2 Changelog

• 250 skill playbooks — full coverage: vuln classes, protocols, frameworks, post-exploitation, payloads
• 109 security tools — netexec, certipy, sliver, ligolo-ng, chisel, dalfox, xsstrike, wifite, pacu, prowler and more
• Intelligence data layer — attack chains, WAF signatures, CVE correlations, tech correlations loaded at runtime
• Client-side pentest — Playwright browser automation for DOM XSS, auth flows, CSRF PoC, SPA endpoint discovery
• Live catalog — tools-catalog.json fetched from GitHub at startup, always current

---

Quickstart

# Install
bunx oh-my-open-pentest install

# Run a full engagement
fullscan https://target.example.com

# Run against a network range (auto-selects red-team mode)
fullscan 10.0.0.1/24

# CTF target
fullscan hackthebox.eu/machines/SomeMachine

The agent detects the engagement type, selects the matching mode and skill chain, and runs the full cycle autonomously.

---

Installation

OpenCode Plugin (Ultimate Edition)

Requires OpenCode and Bun.

bunx oh-my-open-pentest install

The wizard sets your engagement mode, AI provider, and model assignments, then verifies the installation.

Codex Light Edition

Requires Node.js (no Bun needed).

npx lazycodex-ai install

Register the plugin

# Verify plugin is registered
cat ~/.config/opencode/opencode.json | grep plugin
# Expected: "plugin": ["file:///path/to/oh-my-open-pentest/dist/index.js"]

If missing, add it manually:

jq --arg p "file:///path/to/oh-my-open-pentest/dist/index.js" '.plugin = [$p]' \
  ~/.config/opencode/opencode.json > /tmp/oc.json && mv /tmp/oc.json ~/.config/opencode/opencode.json

Verify

bunx oh-my-open-pentest doctor

---

Engagement Modes

Ten modes. Each optimized for a different context. Auto-detected from the target, or set explicitly with /mode.

Mode	Use case	Tool priority	Output
auto	Unknown target	Adaptive	Standard
bug-bounty	HackerOne, Bugcrowd, Intigriti	Recon → Enum → Exploit	HackerOne format
red-team	Stealth ops, persistence, AD	Recon → Exploit → Enum	Executive summary
ctf	HackTheBox, TryHackMe, picoCTF	Exploit → Enum → Recon	Flag submission
blue-team	Detection, IR, defensive audit	Enum → Recon → Report	IR report
offensive	Aggressive exploitation, PoC chains	Exploit → Enum → Recon	Technical
grey-hat	Balanced	Balanced	Technical
forensic	Evidence preservation, disk/memory	Forensics → Report	Chain-of-custody
reverse-engineering	Binaries, firmware, malware	RE → Exploit → Utility	Technical RE
mobile-pentest	Android/iOS app assessment	Mobile → Enum → Exploit	OWASP Mobile Top 10

/mode red-team
/mode bug-bounty
/mode ctf

---

Skill Library

250 skill playbooks — executable by the agent, readable by you.

Vulnerability Classes

/vuln-sqli          /vuln-xss           /vuln-ssrf          /vuln-cors
/vuln-idor          /vuln-rce           /vuln-xxe           /vuln-ssti
/vuln-deserialization                   /vuln-file-upload   /vuln-http-smuggling
/vuln-race-conditions                   /vuln-business-logic /vuln-csrf
/vuln-jwt           /vuln-oauth         /vuln-2fa-bypass    /vuln-account-takeover
/vuln-bfla          /vuln-websocket     /vuln-grpc          /vuln-waf-bypass
/vuln-subdomain-takeover                /vuln-prototype-pollution

Reconnaissance

/recon-full         /recon-subdomain    /recon-internal     /recon-dorking
/recon-js-analysis  /recon-secrets      /recon-shodan       /recon-asn-whois

Post-Exploitation

/post-linux-privesc         /post-windows-privesc       /post-pivoting
/post-lateral-movement      /post-credential-dumping    /post-bloodhound
/post-container-escape

Payload Collections

/payload-xss    /payload-sqli   /payload-ssrf   /payload-ssti
/payload-xxe    /payload-lfi    /payload-command-injection

Technology-Specific

/tech-spring    /tech-wordpress     /tech-docker        /tech-redis
/tech-jenkins   /tech-mongodb       /tech-kubernetes    /tech-firebase

Frameworks

/framework-django   /framework-laravel  /framework-rails    /framework-spring
/framework-express  /framework-nextjs   /framework-fastapi  /framework-dotnet

Protocols

/proto-smb      /proto-kerberos     /proto-graphql      /proto-ssh
/proto-ldap     /proto-rdp          /proto-smtp         /proto-dns

Client-Side

/pentest-browser    — DOM XSS, auth flows, CSRF PoC, SPA endpoints, clickjacking

All skills live in .agents/skills/.

---

Tool Coverage

109 security tools. Missing tools install automatically before first use.

Phase	Tools
Recon / OSINT	subfinder · amass · assetfinder · httpx · naabu · massdns · nmap · masscan · rustscan · theHarvester · sublist3r · spiderfoot · sherlock · holehe · maigret · dnstwist
Secrets	trufflehog · gitleaks · secretfinder
Enumeration	nuclei · ffuf · gobuster · feroxbuster · dirsearch · whatweb · wafw00f · nikto · katana · arjun · testssl · trivy · gospider
Web Exploitation	sqlmap · commix · dalfox · xsstrike · nosqlmap · burpsuite · owasp-zap · mitmproxy
Credential / Auth	hydra · hashcat · john · certipy · kerbrute
Active Directory	netexec · evil-winrm · bloodhound · crackmapexec · impacket · responder · mimikatz
C2 / Pivoting	metasploit · sliver · havoc · mythic · pwncat-cs · chisel · ligolo-ng
Post-Exploitation	peass-ng · pwntools
Wireless	wifite · airgeddon · wifiphisher · bettercap · hcxdumptool
Cloud	pacu · routersploit · prowler · scoutsuite
Phishing	evilginx3 · setoolkit
Forensics / IR	volatility3 · autopsy · binwalk · foremost · bulk_extractor · exiftool · tcpdump · tshark · wireshark · yara · steghide · stegcracker · pspy
Reverse Engineering	ghidra · radare2 · cutter · gdb · pwndbg · ltrace · strace · angr
Mobile	apktool · jadx · frida · objection · mobsf · adb · androguard
Utility	curl · jq · anew · notify · haiti

Full catalog: tools-catalog.json — fetched live from GitHub at startup.

---

Intelligence Data Layer

Structured reference data loaded at runtime for smarter decisions. Sourced from airecon.

File	Contents
attack_chains.json	40+ multi-stage exploitation pathways
vuln_ontology.json	12 vuln categories, CWE mapping, regex signals
waf_signatures.json	WAF fingerprinting signatures
waff_bypass.json	WAF bypass technique library
fuzzer_data.json	Fuzzing payload collections
tech_correlations.json	Technology → known vulnerability correlations
port_correlations.json	Port → service → attack vector mapping
cve_correlations.json	CVE to technology and attack pattern mapping
patterns.json	Detection patterns for vulnerability identification
endpoint_patterns.json	API endpoint pattern library

---

Scope Enforcement

Talos parses the program scope before any active testing. Every target is validated. Out-of-scope assets are refused — even when they appear in the attack path.

Control	Bug Bounty	Red Team	CTF	Offensive
Scope enforcement	Strict	Moderate	Off	Off
DoS protection	On	On	Off	Off
Stealth mode	Off	On	Off	Off
Exfiltration guard	On	On	Off	Off
Auto-stop on violation	On	On	Off	Off

---

Team Mode

Parallel specialist agents for large engagements.

// .opencode/oh-my-open-pentest.jsonc
{
  "team_mode": {
    "enabled": true,
    "max_parallel_members": 4,
    "tmux_visualization": true
  }
}

Built-in team skills:

• hyperplan — 5 adversarial critics review the engagement plan before a tool fires
• security-research — 3 hunters + 2 PoC engineers audit in parallel

---

Uninstallation

jq '.plugin = [.plugin[] | select(. != "oh-my-open-pentest")]' \
    ~/.config/opencode/opencode.json > /tmp/oc.json && \
    mv /tmp/oc.json ~/.config/opencode/opencode.json

rm -f ~/.config/opencode/oh-my-open-pentest.jsonc \
      .opencode/oh-my-open-pentest.jsonc

---

Further Reading

• Manifesto — the philosophy
• Engagement Workflow — full lifecycle
• Engagement Modes — 10 modes in detail
• Tool Reference — 109 tools by phase
• Installation Guide — step-by-step setup
• Team Mode — parallel agent coordination

---

Contributions welcome. PRs to dev.