Skill UI

Guides security teams through enabling Microsoft Defender for Cloud plans, configuring multi-cloud connectors, analyzing Secure Score recommendations, and automating security controls in Azure to protect workloads across servers, containers, storage, and databases.

Monitors and maps threat actor assets by combining Shodan/Censys scans, passive DNS, certificate transparency, WHOIS lookups, and fingerprinting to track command-and-control servers, phishing sites, and pivots over time for security teams.

Detects LLM prompt injection attacks by combining regex signatures, heuristic anomaly scoring, and a DeBERTa classifier to pre-filter malicious inputs before they reach language models, useful in security monitoring or pre-processing layers.

Sets up input and output validation guardrails for LLM apps by combining NeMo Guardrails Colang flows, custom Python PII/content validators, and Guardrails AI schema enforcement so inputs, outputs, and downstream systems stay compliant and safe.

Tooling for AI/ML security assessment, detecting prompt injection, jailbreak, model inversion, data poisoning, and agent tool abuse while mapping findings to MITRE ATLAS and recommending guardrails for LLMs, classifiers, and embedding models.