Skill UI

This skill automates the iterative process of academic paper refinement, simulating the rigorous peer-review cycle of top-tier conferences (e.g., NeurIPS, ICML). It connects with an external LLM (MiniMax) to receive multi-round, critical feedback, tracks state persistence, and guides the user through necessary revisions until the research is deemed submission-ready. Ideal for maximizing the quality of technical writing.

This autonomous skill orchestrates the entire scientific research lifecycle, guiding users from a vague research direction to a polished, review-ready document. The pipeline systematically executes Idea Discovery (Workflow 1), full-scale Implementation, Experiment Deployment (Workflow 2), automated Review Loops, and optional Paper Writing (Workflow 3). It handles everything from preliminary literature review and novelty checks to generating high-quality, reproducible code and final academic drafts.

Provides PyTorch idioms and best practices for building reproducible, efficient training loops, model modules, and data pipelines across devices.

Structured workflow that has an AI agent capture failures, diagnose root causes, apply minimal recovery steps, and report introspection to avoid repeat loops or context drift.

Provides patterns and techniques for building robust evaluation loops for AI agents. This skill guides the implementation of self-critique, evaluator-optimizer pipelines, and test-driven refinement workflows. Use it to significantly enhance the accuracy and quality of complex outputs like code, reports, and analyses by implementing structured, iterative improvement cycles (Generate → Evaluate → Critique → Refine).

This advanced guide covers core Anthropic workflow patterns, specifically focusing on tool use (function calling) and building complex, agentic loops. It provides detailed instructions on defining tools using JSON Schema, executing tool calls, and managing the iterative process required for building sophisticated, AI-powered agents using the Anthropic API.

Learn how to build complex, multi-step AI agents using Cohere's Chat API v2. This skill demonstrates defining external tools (like weather APIs or database searches) and implementing the full agent loop: model calls tool -> execute tool -> feed result back -> final answer. Ideal for creating autonomous, tool-using workflows that interact with external systems.

A robust framework for building complex, long-horizon AI agents using LangGraph. It addresses common pitfalls in advanced agent design, such as unbounded state growth (Virtual FS cleanup) and subagent persona leakage (explicit message overriding). Ideal for tasks requiring multi-step planning, task delegation, and self-correction through reflection loops.

Implement robust human-in-the-loop (HITL) approval gates for sophisticated LangGraph workflows. This skill enables the graph to pause execution before critical, irreversible actions (like sending emails or calling tools). It manages complex state serialization, provides clear resume semantics using `Command(resume=...)`, and integrates with external UIs (e.g., Slack) to await human decision, ensuring safe and controlled workflow execution.

This skill provides a development loop for RunwayML, an advanced creative AI platform. It allows users to programmatically generate high-quality videos from images or text prompts using the Python SDK. Ideal for integrating Runway's capabilities into local development workflows, enabling advanced creative video production, editing, and media manipulation.