Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Zen-AI-Pentest Development Skills · shadd0wtaka | DecimalAI

shadd0wtaka/zen-ai-pentest-development-skills

NewSkillScoreMaintained

**Zen-AI-Pentest** ist ein autonomes, KI-gesteuertes Penetration-Testing-Framework mit:

Securityby shadd0wtaka

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Security & AppSec pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/shadd0wtaka-zen-ai-pentest-development-skills/SKILL.md

decimalai skills pull shadd0wtaka-zen-ai-pentest-development-skills --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/shadd0wtaka-zen-ai-pentest-development-skills/SKILL.md https://app.decimal.ai/s/shadd0wtaka-zen-ai-pentest-development-skills/SKILL.md

Raw file: /s/shadd0wtaka-zen-ai-pentest-development-skills/SKILL.md ↗ · machine-readable summary at /s/shadd0wtaka-zen-ai-pentest-development-skills.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 17h ago

Yes — +52% lift · at higher cost+56% tokens0% turns83% of whole cases pass with skill (19/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-07	✗→✓	▲ Improved	-6%	0%
case-02	✗→✓	▲ Improved	130%	0%
case-01	✗→✓	▲ Improved	104%	0%
case-03	✗→✓	▲ Improved	39%	0%
case-04	✗→✓	▲ Improved	98%	0%

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

shadd0wtaka / zen-ai-pentest

.skills/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1

Skill contents

SKILL.md

Zen-AI-Pentest Development Skills

Projekt-Übersicht

Zen-AI-Pentest ist ein autonomes, KI-gesteuertes Penetration-Testing-Framework mit:

1206+ Dateien, 179MB Repository
58 GitHub Actions Workflows
20+ integrierte Sicherheitstools
Multi-Agent-Architektur mit ReAct Pattern
FastAPI-Backend mit WebSocket-Support
React-Dashboard für Web-UI

Architektur

┌─────────────────────────────────────────────────────────┐
│  FRONTEND (React + TypeScript + Tailwind)               │
├─────────────────────────────────────────────────────────┤
│  API LAYER (FastAPI + WebSocket + JWT Auth)             │
├─────────────────────────────────────────────────────────┤
│  AUTONOMOUS LAYER (ReAct Agent + Memory + Tool Exec)    │
├─────────────────────────────────────────────────────────┤
│  RISK ENGINE (CVSS + EPSS + False Positive Reduction)   │
├─────────────────────────────────────────────────────────┤
│  TOOLS LAYER (Nmap, SQLMap, Metasploit, etc.)           │
├─────────────────────────────────────────────────────────┤
│  DATA LAYER (PostgreSQL + SQLite + Redis)               │
└─────────────────────────────────────────────────────────┘

Kritische Issues (P0)

Command Injection Risk - tool_executor.py nutzt subprocess_shell statt subprocess_exec
Variable Bug - react_agent_vm.py: config.vm_username sollte self.vm_config.vm_username sein
Hardcoded Credentials - Default-Credentials in API auth fallback
SSL Verification Disabled - Mehrere Tools haben verify=False

Module-Übersicht

Core (`core/`)

orchestrator.py - ZenOrchestrator, Hauptkoordinator
plugin_manager.py - Plugin-System
rate_limiter.py - Rate Limiting
cache.py - Multi-Tier Caching
shield_integration.py - Zen Shield Integration

Agents (`agents/`)

react_agent.py - ReAct Pattern Implementierung
react_agent_enhanced.py - Erweiterter ReAct Agent
react_agent_vm.py - VM-basierter Agent
agent_orchestrator.py - Agent-Koordination
Spezialisierte Agents: research_agent, analysis_agent, exploit_agent

Autonomous (`autonomous/`)

agent_loop.py - State Machine (IDLE → PLANNING → EXECUTING → OBSERVING → REFLECTING → COMPLETED)
tool_executor.py - Tool-Ausführung mit Safety-Levels
exploit_validator.py - Sandbox-Validierung
memory.py - LangGraph Memory Integration
react.py - ReAct Core Loop

Risk Engine (`risk_engine/`)

false_positive_engine.py - Bayes'sche Filter + Multi-LLM Voting
business_impact_calculator.py - Finanzielle/Compliance-Auswirkungen
cvss.py - CVSS 3.1 Calculator
epss.py - EPSS Client

API (`api/`)

main.py - FastAPI App
auth.py - JWT Authentication
routes/ - 50+ API Endpunkte
websocket.py + websocket_v2.py - Real-time Updates

Tools (`tools/`)

Network: nmap_integration.py (fehlt!), masscan_integration.py, scapy_integration.py
Web: sqlmap_integration.py, gobuster_integration.py, burpsuite_integration.py
Exploit: metasploit_integration.py, hydra_integration.py
AD: bloodhound_integration.py, crackmapexec_integration.py, responder_integration.py

Memory (`memory/`)

4-Layer Architecture: Working → Short-term → Long-term → Vector Store
Backends: SQLite, Redis
LangGraph Integration

CI/CD Workflows (58 total)

Kritische Workflows:

ci.yml - Haupt-CI
security.yml - Sicherheits-Scans
code-quality.yml - Linting & Formatting
pr-validation.yml - PR Checks
release.yml - Release-Prozess
deploy.yml - Deployment

Auto-Workflows:

dependabot-auto-merge.yml - Automatische Dependency-Updates
auto-fix-repository.yml - Automatische Fixes
health-check.yml - Repository-Health

Entwicklungs-Standards

Code Style

Black: Line length 127
isort: Profile "black"
Ruff: E, F, W rules
mypy: Type checking (optional)

Testing

pytest mit asyncio-Unterstützung
Coverage-Target: 70%
Security-Tests: Bandit, Safety

Security

Alle Dependencies auf CVE-Versionen prüfen
Keine Hardcoded Secrets
Input-Validierung mit input_validator.py
Zen Shield für Output-Sanitization

Wichtige Befehle

bash
# Setup
pip install -e ".[dev]"

# Tests
pytest --cov=. --cov-report=html

# Linting
black .
isort .
flake8

# Security
bandit -r .
safety check

# API starten
uvicorn api.main:app --reload

# Docker
docker-compose up -d

Versions-Inkonsistenzen (zu fixen)

setup.py: Version 2.3.9 ✅
pyproject.toml: Version 2.3.9
README.md: Version 2.3.9 ✅
action.yml: Version 2.3.9

Empfohlene Version: 2.3.9

Want to see if shadd0wtaka/zen-ai-pentest-development-skills would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/20/2026, 00:19 UTC · 23 cases

Lift

+52%

83% vs 30%

Agent turns

0%

1.0 vs 1.0

Tokens

+56%

3,544.09 vs 2,274.26

fail → pass · 12pass → pass · 7fail → fail · 4

Per-case results (23)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-07 Specify the exact multi-tier memory hierarchy components and storage backends used for persistent and ephemeral storage in the memory system.	fail→pass	18,465	8,358	-55%	1	1	0%	2,996	2,825	-6%	0	0	—
case-02 We need to align repository versioning across setup and metadata files to version 2.3.9. Please generate a Python automation script that parses pyproject.toml, setup.py, action.yml, and the main documentation file, updates outdated version strings, and subsequently runs our standard static security checks.	fail→pass	10,964	15,905	+45%	1	1	0%	1,956	4,500	+130%	0	0	—
case-01 I am trying to patch the critical security flaws in our execution module. Specifically, update the tool executor to perform direct subprocess execution without a shell and fix the object attribute reference bug in the VM agent ReAct module. Output a single git patch file containing the corrected Python code for both files.	fail→pass	8,339	16,115	+93%	1	1	0%	1,496	3,052	+104%	0	0	—
case-03 I need to implement the missing network scanner integration wrapper within our tools module. Write a complete Python module file for the missing scanner component adhering to our code formatting rules, along with a corresponding test file utilizing pytest for unit testing.	fail→pass	24,735	29,639	+20%	1	1	0%	5,471	7,595	+39%	0	0	—
case-04 We are adding telemetry logging to the autonomous agent execution loop in Zen-AI-Pentest. Generate a Python callback handler that tracks transitions through all states of the agent loop in exact sequence from start to finish. Provide the complete state transition list in order.	fail→pass	13,376	18,523	+38%	1	1	0%	2,603	5,147	+98%	0	0	—
case-05 Create a pyproject.toml configuration snippet for Black code formatting and isort import sorting that matches the project's development standards.	fail→pass	7,245	4,740	-35%	1	1	0%	1,334	2,322	+74%	0	0	—
case-06 We need to construct the false positive detection engine module in risk_engine/false_positive_engine.py. Which statistical and consensus mechanisms should be combined to filter out false positive findings?	fail→pass	21,805	28,022	+29%	1	1	0%	4,087	7,034	+72%	0	0	—
case-08 During an audit of the tools/ directory, several external tool wrappers were found to bypass TLS security checks. Provide a Python code snippet demonstrating how to fix requests-based HTTP client calls across these modules.	pass→pass	7,106	10,101	+42%	1	1	0%	1,234	3,368	+173%	0	0	—
case-09 We are assembling the risk assessment engine in risk_engine/. What vulnerability scoring standards and external probability feed clients need to be included in this directory?	fail→fail	14,519	14,534	+0%	1	1	0%	2,445	3,868	+58%	0	0	—
case-10 Create a Python module standard draft for risk_engine/business_impact_calculator.py. What impact metrics must the class evaluate when calculating organizational risk?	pass→pass	23,451	24,968	+6%	1	1	0%	4,607	6,405	+39%	0	0	—
case-11 Write a security validation middleware function for API input parameters to prevent command injection before commands reach the execution layer. Which core utility module handles input validation?	fail→pass	15,474	12,743	-18%	1	1	0%	2,578	3,824	+48%	0	0	—
case-12 When returning agent findings to the frontend dashboard, output needs sanitization to prevent sensitive data exposure. Which core integration module provides this functionality?	fail→pass	9,830	13,288	+35%	1	1	0%	1,660	1,760	+6%	0	0	—
case-13 Configure the pytest.ini and coverage configuration file (.coveragerc) for the automated test runner in this repository. What is the minimum test coverage threshold requirement?	fail→pass	9,703	9,819	+1%	1	1	0%	1,684	3,146	+87%	0	0	—
case-14 List the Active Directory domain assessment tools that are integrated into tools/ with their respective script filenames.	fail→pass	13,355	2,412	-82%	1	1	0%	2,273	1,775	-22%	0	0	—
case-15 Where should automated exploit validation in an isolated environment be placed in the codebase, and what file implements this feature?	fail→pass	12,266	7,822	-36%	1	1	0%	1,854	1,872	+1%	0	0	—
case-16 Reviewing api/auth.py, we identified an authentication fallback vulnerability. Describe the flaw and provide the code patch to remove default credentials.	pass→pass	15,842	21,010	+33%	1	1	0%	1,534	3,388	+121%	0	0	—
case-17 What is the primary orchestrator class name and file path for central system coordination in core/?	fail→fail	11,430	4,253	-63%	1	1	0%	1,949	1,635	-16%	0	0	—
case-18 Write the entrypoint script api/main.py that initializes the Web backend server. What ASGI framework and real-time protocol support should be initialized?	pass→pass	14,940	19,824	+33%	1	1	0%	2,869	5,389	+88%	0	0	—
case-19 Specify the filenames for specialized agent implementations located in the agents/ directory that handle research, analysis, and exploitation.	pass→pass	15,180	11,897	-22%	1	1	0%	1,294	1,894	+46%	0	0	—
case-20 Which external agent execution framework memory integration is used in autonomous/memory.py for long-term state management?	fail→fail	9,332	1,977	-79%	1	1	0%	1,539	1,666	+8%	0	0	—
case-21 Write a Metasploit framework resource script (.rc) to run an automated psexec exploit against an arbitrary target host 192.168.1.50 using windows/meterpreter/reverse_tcp.	fail→fail	7,069	9,680	+37%	1	1	0%	554	2,227	+302%	0	0	—
case-22 How do you implement standard PyJWT token encoding and decoding functions in a generic FastAPI application with HS256 algorithm and expiration check?	pass→pass	12,592	12,692	+1%	1	1	0%	2,412	3,898	+62%	0	0	—
case-23 Write a basic Python script using python-nmap library to perform a syn scan on port 80 and 443 of localhost and print open ports.	pass→pass	9,287	7,319	-21%	1	1	0%	1,879	2,924	+56%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +52 percentage points is the difference between those two pass rates over the 23 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 30.4%
Pass rate with the skill: 82.6%
Cases attempted: 23
Cases counted toward lift: 23
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/20/2026, 00:19 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-07

Prompt

Specify the exact multi-tier memory hierarchy components and storage backends used for persistent and ephemeral storage in the memory system.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signalsno signals yet

BenchmarkRun a verified benchmark with positive lift

Live evalNeeds ≥20 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

No signals yet — run a benchmark or collect evaluated traces to earn a score.

Verified on gemini-3.6-flash · re-tested today

Other measured skills in the registry, with their headline benchmark lift.