Skill UI

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.

A comprehensive tool and API wrapper for accessing, parsing, and analyzing research papers published on Hugging Face and arXiv. Users can input various URLs or IDs (e.g., arXiv ID) to retrieve structured metadata, full markdown content, and automatically link associated resources such as models, datasets, and application spaces. Ideal for researchers and developers needing structured access to cutting-edge AI knowledge.

Assesses AI/ML systems for prompt injection, jailbreak, model inversion, data poisoning, and agent tool abuse using MITRE ATLAS mapping, signature detection, and scoring metrics before deployment.

Describes how to use the MemOS local memory toolkit to recall past conversations, search shared histories, and choose the right memory tool when automatic recall is insufficient.

This skill enables the generation of high-quality, cinematic videos using a highly structured prompt format (JSON). Users define detailed cinematic elements—including background, characters, camera movements, dialogue, and audio—and can optionally provide a reference image to guide the visual style or content of the final output. It automates the complex AIGC video generation workflow.

A powerful tool leveraging OpenRouter and the Gemini 3 Pro Image model to create, modify, and composite high-quality images. Supports three primary use cases: prompt-only generation from text, detailed single-image edits (inpainting/outpainting), and complex multi-image composition. Users can specify resolutions up to 4K, making it suitable for professional creative and design workflows.

This skill provides comprehensive tools to download, export, and deeply inspect existing Arize traces and spans. It is essential for understanding the full execution flow of complex LLM applications, debugging runtime errors, and performing root-cause analysis. Users can retrieve detailed data filtered by Trace ID, Span ID, or Session ID, enabling thorough analysis of conversational or multi-step agent behaviors.

Reads `.agent.md` or `.prompt.md` files, assesses task complexity, tooling, and requirements, then recommends the most cost-efficient GitHub Copilot model(s) (e.g., GPT-5, Claude Sonnet) with rationale for users across different subscription tiers and priorities.

The Phoenix CLI is a powerful tool for debugging complex LLM and agent workflows. It allows users to fetch detailed traces, spans, and sessions, analyze errors, and structure findings using open and axial coding techniques. It is essential for investigating complex LLM failures, building robust failure taxonomies, and guiding the next steps in application evaluation after initial instrumentation.

The Skillshare CLI provides a centralized system for managing, syncing, and deploying AI skills and agents across numerous AI tools (like Claude, Cursor, etc.). It handles global and project-specific configurations, allowing users to install, update, audit, and coordinate skills and custom extras (rules, commands) from a single source of truth.

DiffDock is a cutting-edge deep learning tool leveraging diffusion models to predict the 3D binding poses of small molecule ligands to protein targets. It supports structure-based drug discovery, enabling single-complex docking, batch virtual screening, and lead optimization. Users can input protein structures (PDB) or sequences, along with diverse ligand formats (SMILES, SDF, MOL2). The tool outputs predicted poses and confidence scores, making it crucial for identifying potential drug candidates, though it does not predict binding affinity.

The ESM skill provides comprehensive tools for computational biology, leveraging ESM3 and ESM-C models. It enables sophisticated protein sequence generation, structure prediction (sequence to 3D structure), inverse folding (structure to sequence design), and generating function-conditioned proteins. Use it for de novo protein design, feature embedding, and advanced structural analysis in bioinformatics workflows.