aiml-code-vuln-benchmark

To generate OWASP vulnerability samples for security benchmarking, this Skill creates concrete vulnerable code and PoC exploits across six classes. It enforces minimum code length and vulnerability sinks without placeholders to anchor automated evaluation data.

Supported OWASP vulnerability classes include sql_injection, xss_stored, rce_command_injection, path_traversal, ssrf, and insecure_deserialization. The generator targets these specific classes to produce anchored vulnerability data for testing static analyzers and red-teaming tooling.

To create proof-of-concept exploits for testing static analyzers, the Skill generates complete vulnerability samples containing concrete vulnerable code and working PoC payloads. It enforces minimum code length and specific vulnerability sinks to validate automated security analyzers.

The generated vulnerability code does not contain placeholders or abstract patterns. It enforces concrete vulnerable code and proof-of-concept exploits with specific vulnerability sinks to ensure actionable benchmark data for automated evaluation pipelines.

You can use this to benchmark AI-generated exploit payloads for path traversal, as it generates complete vulnerability samples for path_traversal and five other OWASP classes. The outputs provide anchored vulnerability data for evaluating model-generated exploit code and PoC payloads.