Skill UI

This skill guides security analysts through advanced threat-hunting methodologies focused on detecting DNS-based persistence. It analyzes passive DNS history, zone file audits, and API data (like SecurityTrails) to identify unauthorized record modifications, subdomain takeovers, and DNS hijacking attempts that bypass traditional security controls. Ideal for SOC analysts and threat hunters.

This comprehensive skill automates the entire process of generating novel, robust, and benchmark-grounded research ideas in robotics and embodied AI. It guides users from a broad initial concept to concrete, actionable project proposals by chaining rigorous literature surveys, novelty verification, and critical reviews. The workflow enforces robotics constraints, prioritizing simulation-first, falsifiable, and feasible real-world applications.

This skill guides the implementation of network traffic baselining using historical NetFlow/IPFIX data. It leverages Python's pandas library to compute detailed statistical profiles, including hourly/daily distributions, per-host metrics, and protocol ratios. Anomalies are robustly detected using z-score and Interquartile Range (IQR) methods, assisting SOC analysts in identifying unusual activities like data exfiltration or beaconing.

Transforms vague research directions into highly concrete, elegant, and frontier-aware methodological plans. This skill guides users through an iterative, multi-phase refinement process, ensuring the final output maintains a strong Problem Anchor, focuses on minimal adequate contributions, and is ready for top-tier academic publication.

This skill provides a multi-round, deep critical review of research proposals, papers, or experimental results by simulating the perspective of a senior ML reviewer (e.g., NeurIPS/ICML level). It guides the user through identifying logical gaps, missing experiments, and narrative weaknesses, ensuring high fidelity and iterative refinement of technical work.

This skill automates the end-to-end workflow for deploying and executing ML training and inference experiments. It intelligently detects the target environment (Local GPU, Remote SSH, Vast.ai, Modal serverless) and handles necessary steps like GPU pre-checks, code synchronization (rsync/git), integrating ML tracking tools (W&B), and launching the compute job via screen sessions or serverless deployment. Use when initiating a complex training job.

This tool audits the token consumption across all components (agents, skills, rules, MCP servers) within a Claude Code session. It analyzes token overhead, identifies bloat (e.g., overly long descriptions, redundant components), and generates a detailed context budget report. Use it to maintain optimal context window usage, improve performance, and prevent context saturation.

This skill provides guidance on implementing common architectural patterns and best practices within the PyTorch framework. It is designed for developers building sophisticated deep learning models, helping to improve structure, efficiency, and code maintainability when working with neural networks.

This skill systematically probes an AI model's core behavioral patterns across six critical dimensions, including reasoning, refusal, formatting, and grounding. It generates a comprehensive, styled HTML report with radar charts and actionable insights, allowing users to understand a model's true capabilities, compare different models, or perform compliance and red-team assessments without needing an external API key.

This autonomous skill factory converts raw, ambiguous human input—such as messy transcripts, documents, screenshots, or vague workflows—into complete, production-ready, cross-platform agent skills. It acts as a powerful intermediary, deriving complex intent and automating the creation of reusable software, bridging the gap between human articulation and technical specifications. Ideal for automating repetitive tasks and building custom agent capabilities with minimal coding knowledge.

This skill detects AI-generated deepfake audio used in voice phishing (vishing) attacks. It functions by extracting critical spectral features (including MFCCs, spectral centroid, and zero-crossing rate) and passing them through advanced machine learning models (like Random Forest and Gradient Boosting). The system supports batch forensic analysis, provides confidence scores, and generates detailed reports to verify the authenticity of recorded voice calls, making it essential for cybersecurity incident response and fraud investigation.

This skill is designed for performing local, on-premise evaluations and benchmarking of models hosted on the Hugging Face Hub. It facilitates running specialized evaluation frameworks like `inspect-ai` and `lighteval` against local hardware. Users can select optimal inference backends, including `vllm` for high throughput, or fall back to Hugging Face Transformers or `accelerate`, ensuring comprehensive smoke testing and task selection for LLMs. It focuses purely on the local execution process, separate from remote jobs or publishing workflows.