CRSBench

Generate CRSBench experiment config YAML files conversationally. Use this skill whenever the user mentions experiment configs, wants to set up a CRSBench run, asks to create or scaffold a YAML config, describes an experiment they want to run (mentioning CRS names, benchmark suites, bugfinding/bugfixing), or needs help writing configuration for crsbench. Even partial descriptions like "run crs-libfuzzer on sanity" should trigger this skill.

Analyze CRSBench experiment-data directories to diagnose issues. Use when reviewing crsbench run results, identifying failures, and summarizing LLM costs.

Analyze vulnerabilities from crash logs, POV files, patches, and source code. Use when analyzing CPV vulnerabilities, preparing vuln.yaml generation, or investigating security issues in CRSBench benchmarks.

Generate vuln.yaml files from vulnerability analysis results. Use when creating or updating vuln.yaml for CRSBench CPVs based on vulnerability analysis markdown.