"What's our vector, Victor?" - A collection of intentionally vulnerable code examples for security testing, training, and educational purposes.
DO NOT deploy these applications. These are sample code snippets that have not been tested as executables.
This repository contains deliberately insecure code designed to demonstrate common security vulnerabilities. All examples are for educational and authorized security testing purposes only. Validation of these examples is still a work-in-progress, these are generated by AI and my be slop examples, tbd, until further review.
Example agentic LLM application demonstrating OWASP Top 10 for LLM (2025) vulnerabilities:
- agent.py - Main vulnerable agent with comprehensive security issues
- prompt_injection_examples.py - LLM01: Direct/indirect prompt injection and jailbreaks
- sensitive_data_exposure.py - LLM02: Hardcoded secrets, PII leakage, system prompt exposure
- excessive_agency.py - LLM06: Unrestricted capabilities and auto-execution
- config.yaml - Vulnerable configuration with hardcoded credentials
- requirements.txt - Python dependencies
OWASP Top 10 for LLM Coverage:
- ✓ LLM01: Prompt Injection - No input validation, direct concatenation
- ✓ LLM02: Sensitive Information Disclosure - Hardcoded secrets, PII exposure
- ✓ LLM03: Supply Chain Vulnerabilities - No model verification, untrusted sources
- ✓ LLM04: Data and Model Poisoning - Insecure pickle, no data validation
- ✓ LLM05: Improper Output Handling - Auto-executing commands, eval() usage
- ✓ LLM06: Excessive Agency - Unrestricted file/DB/API access, no approvals
- ✓ LLM07: System Prompt Leakage - Weak protection, credentials in prompts
- ✓ LLM08: Vector and Embedding Weaknesses - No validation, poisoning risk
- ✓ LLM09: Misinformation - No fact-checking or source attribution
- ✓ LLM10: Unbounded Consumption - No rate limits or resource constraints
References:
- OWASP Top 10 for LLM Applications 2025
- OWASP GenAI Security Project
- NIST AI Risk Management Framework
Examples of insecure Docker configurations and container practices:
- Dockerfile.python-vulnerable - Python app with multiple security issues
- Dockerfile.nodejs-vulnerable - Node.js app with vulnerabilities
- Dockerfile.java-vulnerable - Java app with security flaws
- Dockerfile.secrets-exposed - Hardcoded secrets and credentials
- Dockerfile.rootful-privileged - Privileged containers running as root
- Dockerfile.multistage-bad - Insecure multi-stage builds
- docker-compose.vulnerable.yml - Insecure Docker Compose configuration
Infrastructure-as-Code examples with security misconfigurations:
- aws_s3_vulnerable.tf - Publicly accessible S3 buckets, weak encryption
- aws_ec2_vulnerable.tf - Insecure EC2 instances, security groups, SSH keys
- aws_rds_vulnerable.tf - Unencrypted databases, weak passwords, public access
- aws_iam_vulnerable.tf - Overly permissive IAM policies and roles
- aws_misc_vulnerable.tf - Additional AWS security issues
Python web application examples demonstrating OWASP Top 10 (2021) vulnerabilities:
- SQL Injection
- Cross-Site Scripting (XSS)
- XML External Entity (XXE)
- Command Injection
- Path Traversal
- Server-Side Request Forgery (SSRF)
- Insecure Deserialization
- Broken Authentication
- Weak Cryptography
- Hardcoded Secrets
Each directory contains its own README with specific vulnerability descriptions.
# Clone the repository
git clone https://github.com/levinebw/vectorvictor.git
cd vectorvictor
# Navigate to specific vulnerability examples
cd vulnerable_llm_agent
pip install -r requirements.txt
python agent.py- OWASP Top 10
- OWASP Top 10 for LLM Applications
- CWE - Common Weakness Enumeration
- MITRE ATT&CK Framework
This repository is for educational and demo purposes only.
Contributions of additional vulnerability examples are welcome. Please ensure:
- Code is clearly documented
- Vulnerabilities are explicitly noted
- Examples are realistic and educational
- No actual sensitive data is included
This project is provided "as-is" for educational and demonstration purposes. See LICENSE file for details.