AI-Skills-Collections

Maintain this specific academic website repo safely and consistently. Use this skill whenever the user asks to update, review, polish, reorganize, or extend this website, especially when the request touches homepage identity statements, publications, CV assets, shared templates, or duplicated academic content.

Delegate coding tasks to Claude Code (Anthropic's CLI agent). Use for building features, refactoring, PR reviews, and iterative coding. Requires the claude CLI installed.

Delegate coding tasks to OpenAI Codex CLI agent. Use for building features, refactoring, PR reviews, and batch issue fixing. Requires the codex CLI and a git repository.

Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.

Complete guide to using and extending Hermes Agent — CLI usage, setup, configuration, spawning additional agents, gateway platforms, skills, voice, tools, profiles, and a concise contributor reference. Load this skill when helping users configure Hermes, troubleshoot issues, spawn agent instances, or make code contributions.

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

Delegate coding tasks to OpenCode CLI agent for feature implementation, refactoring, PR review, and long-running autonomous sessions. Requires the opencode CLI installed and authenticated.

Configure Zed editor settings, keybindings, agent profiles, and AI integrations. Use when modifying Zed configuration, setting up agent profiles such as plan mode, configuring language servers, themes, or AI models. Keywords: zed, editor, settings, keybindings, keymap, LSP, language server, theme, agent, configure, model, MiniMax.

PyTorch deep learning framework. Use for building neural networks, tensor operations, automatic differentiation, GPU acceleration, and custom model training. Best for custom deep learning models, research prototyping, and production deployment. For pre-trained transformers use transformers; for distributed training use deepspeed or accelerate.

BitsAndBytes for quantization of LLMs. Use for 8-bit and 4-bit model quantization, mixed INT8 quantization, NF4 quantization, and memory-efficient LLM loading. Best for running large models on limited GPU memory. For model loading use transformers; for efficient fine-tuning use peft; for distributed training use deepspeed.

AI agent that autonomously interacts with the web via Chromium/CDP. Makes websites accessible for LLMs.

Check Cursor usage, billing, and quota through the shared provider usage checker. Reuse local browser session cookies and call Cursor's authenticated usage APIs before considering any browser automation.

Hugging Face Hub for model and dataset management. Use for downloading/uploading models, managing model repositories, browsing model cards, version control for models, and accessing datasets. Best for model management, sharing models, and downloading pre-trained assets. For model loading use transformers; for datasets use datasets library.

Use the MiniMax `mmx` CLI as a unified multimodal generation and utility tool. Use this skill whenever the user asks to use MiniMax, mmx, MiniMax CLI, or wants terminal-driven text chat, image generation, video generation, text-to-speech, music generation or covers, image understanding, web search, quota checks, or MiniMax authentication/configuration.

PEFT (Parameter-Efficient Fine-Tuning) for LoRA, QLoRA, and adapter methods. Use for fine-tuning large language models efficiently by training only a small number of parameters. Best for fine-tuning LLMs on limited GPU memory. For full fine-tuning use transformers; for RLHF use trl; for quantization use bitsandbytes.

Hugging Face Transformers library. Use for working with pre-trained transformer models (BERT, GPT, Llama, etc.), text classification, named entity recognition, question answering, text generation, translation, summarization, and fine-tuning. Best for NLP tasks and LLM inference. For fine-tuning use peft or trl; for distributed training use accelerate.

TRL (Transformer Reinforcement Learning) for SFT, RLHF, DPO training. Use for supervised fine-tuning, reward modeling, RLHF (PPO), and Direct Preference Optimization (DPO). Best for training instruction-following LLMs. For model loading use transformers; for efficient fine-tuning use peft; for quantization use bitsandbytes.

Hugging Face Accelerate for distributed training and mixed precision. Use for multi-GPU training, TPU support, mixed precision (FP16/BF16), gradient accumulation, and seamless scaling from laptop to cloud. Best for scaling PyTorch training across devices. For deep learning use pytorch; for fine-tuning use peft or trl.

Hugging Face Datasets library for loading and processing datasets. Use for loading datasets from Hub, processing with map/batch, tokenization, train-test splits, caching, and memory-mapped datasets. Best for NLP and ML datasets. For model training use transformers; for loading from files use PyTorch DataLoader.

DeepSpeed for distributed deep learning training. Use for ZeRO optimization, mixed precision training, pipeline parallelism, and memory-efficient training of large models. Best for training very large models (7B+) and optimizing GPU memory. For training use pytorch or pytorch-lightning; for distributed use accelerate.

PyTorch Lightning for simplified training loops. Use for structured deep learning training, reducing boilerplate, built-in logging, multi-GPU/TPU support, and checkpointing. Best for research and production ML training. For deep learning use pytorch; for distributed use deepspeed or accelerate.

Query Ensembl genome database REST API for 250+ species. Gene lookups, sequence retrieval, variant analysis, comparative genomics, orthologs, VEP predictions, for genomic research.

Access NCBI GEO for gene expression/genomics data. Search/download microarray and RNA-seq datasets (GSE, GSM, GPL), retrieve SOFT/Matrix files, for transcriptomics and expression analysis.

Human Cell Atlas data portal. Use for accessing single-cell reference datasets, cell type markers, and standard atlases.

Direct REST API access to KEGG (academic use only). Pathway analysis, gene-pathway mapping, metabolic pathways, drug interactions, ID conversion. For Python workflows with multiple databases, prefer bioservices. Use this for direct HTTP/REST work or KEGG-specific control.

Query Reactome REST API for pathway analysis, enrichment, gene-pathway mapping, disease pathways, molecular interactions, expression analysis, for systems biology studies.

Acts as a peer review assistant to evaluate bioinformatics manuscripts. Identifies missing baselines, methodological flaws, and data leakage.

Drafts highly precise Methods sections for spatial transcriptomics, single-cell RNA-seq (e.g., muscle aging/regeneration), and 3D interpolation pipelines.

Translates complex deep learning architectures (VQ-VAE, Transformers, Neural ODEs) into standard biological manuscript text, focusing on the biological rationale for the architecture.

Multi-omics integration for single-cell data. Use for integrating multiple modalities like RNA, ATAC, and protein in single-cell analysis.

Access NIH Metabolomics Workbench via REST API (4,200+ studies). Query metabolites, RefMet nomenclature, MS/NMR data, m/z searches, study metadata, for metabolomics and biomarker discovery.

Direct REST API access to UniProt. Protein searches, FASTA retrieval, ID mapping, Swiss-Prot/TrEMBL. For Python workflows with multiple databases, prefer bioservices (unified interface to 40+ services). Use this for direct HTTP/REST work or UniProt-specific control.

R package for single-cell ATAC-seq analysis. Use for chromatin accessibility analysis, peak calling, motif enrichment, and integration with gene expression.

R package for chromatin analysis. Use for single-cell ATAC-seq analysis, chromatin accessibility, and integration with Seurat.

Differential gene expression analysis (Python DESeq2). Identify DE genes from bulk RNA-seq counts, Wald tests, FDR correction, volcano/MA plots, for RNA-seq analysis.

Data structure for annotated matrices in single-cell analysis. Use when working with .h5ad files or integrating with the scverse ecosystem. This is the data format skill—for analysis workflows use scanpy; for probabilistic models use scvi-tools; for population-scale queries use cellxgene-census.

Convert between .h5ad (AnnData) files and R single-cell data formats (Seurat, SingleCellExperiment) using anndataR. Use this skill whenever the user mentions: h5ad files, AnnData objects, converting between Python scanpy and R (Seurat/SingleCellExperiment), moving single-cell data between Python and R ecosystems, or needs to read/write .h5ad files from R. This skill provides pure R H5AD interface without Python dependencies.

R package for single-cell trajectory analysis. Use for reconstructing pseudotime trajectories, identifying cell differentiation paths, analyzing branch points, and ordering cells along developmental processes. Best for understanding dynamic biological processes from scRNA-seq data.

Standard single-cell RNA-seq analysis pipeline. Use for QC, normalization, dimensionality reduction (PCA/UMAP/t-SNE), clustering, differential expression, and visualization. Best for exploratory scRNA-seq analysis with established workflows. For deep learning models use scvi-tools; for data format questions use anndata.

R package for converting h5ad files (HDF5 AnnData format from scanpy) to Seurat or SingleCellExperiment objects. Use whenever working with single-cell data in h5ad format that needs to be converted to R. Includes support for Visium spatial transcriptomics data, raw counts, cell metadata, and dimensionality reduction embeddings.