Skill UI

This skill outlines the implementation of advanced runtime defenses for production Large Language Model (LLM) applications. It covers using multiple industry-leading guardrail systems—including Llama Guard, NeMo Guardrails, and LLM Guard—to detect and block adversarial attacks such as prompt injection, jailbreaks, and toxic content. It provides a comprehensive defense-in-depth strategy for ensuring model safety and output integrity in real-world deployments.

This skill guides the deployment and strategic use of honeytokens and canarytokens (decoy credentials, files, URLs, etc.). These high-fidelity decoys are designed to detect unauthorized access, insider threats, or adversary reconnaissance by triggering alerts upon any interaction, providing early warning deep within the kill chain.

Falco is a powerful CNCF project for real-time runtime security monitoring. This guide teaches authors how to write and deploy custom detection rules using Falco's robust YAML rule engine and modern eBPF driver. It is designed to detect critical threats such as container escapes, namespace manipulation, privileged mounts, and anomalous syscalls in Kubernetes and Docker environments, significantly hardening the overall container security posture.

This skill provides comprehensive tools to detect data poisoning and model backdooring attacks throughout the machine learning lifecycle. It checks data provenance, detects label-flip anomalies (using Cleanlab), and analyzes model activations and spectra (using ART) to identify malicious triggers or compromised weights, ensuring ML supply chain integrity.

A comprehensive tool and methodology for identifying and preventing dependency confusion attacks across major package ecosystems (npm, PyPI, Maven). It detects instances where internal, private package names are leaked and subsequently published publicly by attackers. The skill provides both detection (enumerating claimable internal names) and prevention strategies (enforcing registry pinning and scope restrictions) to secure the software supply chain.

This skill provides advanced threat hunting logic for Microsoft Sentinel/Log Analytics. It analyzes both AADGraphActivityLogs and MicrosoftGraphActivityLogs to detect the unique fingerprints and behavioral patterns left by offensive Entra ID tools (like ROADtools, AADInternals, and AzureHound). It detects specific User-Agent strings and characteristic endpoint-sweep patterns, even when tool authors attempt to spoof headers, helping SOC teams identify valid account-based adversary activity (MITRE T1078.004).

A comprehensive security tool designed to detect and defend against subtle prompt injection attacks hidden within untrusted multi-modal content, such as web pages, PDFs, and images. It scans for obfuscated payloads (e.g., hidden CSS, metadata, zero-width characters) by combining content normalization, heuristic analysis, and dedicated ML detectors, ensuring that LLM agents only consume sanitized, trustworthy input.

Provides a comprehensive workflow and tooling (GuardDog) to triage npm packages and dependency trees for signs of malicious activity. It focuses on static analysis of install scripts, detecting credential exfiltration, reverse shells, and worming behavior, thereby safeguarding against software supply-chain attacks.

This skill provides methods and frameworks to detect sophisticated AI attacks, including model stealing, membership inference, and model inversion. It focuses on monitoring query patterns, analyzing confidence exposure, and implementing defensive controls on inference APIs to protect intellectual property and data privacy.

This skill provides a comprehensive, cross-platform workflow for auditing UEFI firmware security. It detects advanced threats like bootkits (e.g., BlackLotus, Bootkitty) and Secure Boot bypasses by verifying the integrity of the platform's trust chain. Key checks include confirming Secure Boot status, assessing the currency and application of DBX revocations, and examining the protection of critical EFI variables using specialized tools. Essential for hardware security assessments and forensic investigation.

This skill detects malicious, misspelled, or brand-jacked package names (typosquatting) across major ecosystems (npm, PyPI, crates.io). It uses advanced techniques like edit-distance and keyboard-proximity analysis, powered by tools like typomania and OSSGadget, to flag potential supply chain attacks *before* a dependency is installed. It is essential for implementing robust CI/CD security gates and proactive dependency screening.

This skill provides expert guidance on design engineering, focusing on UI polish, micro-interactions, and high-craft component decisions. It is ideal for reviewing frontend code (React, CSS, Tailwind) to elevate perceived quality, addressing details like easing, duration, and physical feedback. Use it when subtle detail and refined taste are critical differentiators.