Agent Skills by Hugh
Showing 50 vetted skills indexed across 4 GitHub repositories.
HughYaufable4sci
Transforms scientific research questions into plain and image-rich fables preserving reasoning structure.
HughYauscientific-visualization
Create publication-ready scientific figures with matplotlib, seaborn, and plotly.
HughYaufigure-composer
Compose publication-grade multi-panel figures from a claim and data files.
HughYaulearn
Guides learners through concepts using Socratic tutoring and scaffolded instruction.
HughYauself-awareness
Query Claude Science session metadata, token usage, and execution history via SQLite.
HughYauusing-model-endpoint
Call registered model endpoints over their native HTTP APIs from scoped inference kernels.
HughYaualphafold2
Predicts protein monomer and multimer structures with AlphaFold2 via the ColabFold runner.
HughYausolublempnn
Generates solubility-biased protein sequences from backbone structures using SolubleMPNN inverse folding.
HughYauesmfold2
Predict protein, DNA, RNA, and ligand complex structures with ESMFold2 diffusion co-folding.
HughYauboltz
Predicts protein, nucleic-acid, and small-molecule complex structures with Boltz-2.
HughYauremote-compute-ssh
Orchestrates job submission, monitoring, and result harvesting on SSH and SLURM compute hosts.
HughYauligandmpnn
Design protein sequences around bound ligands, nucleic acids, and metals using LigandMPNN inverse folding.
HughYaualgorithmic-art
Creates generative art with p5.js using seeded randomness and interactive parameter controls.
HughYaupdf-explore
Parses PDFs once into persistent text and images for targeted page navigation and extraction.
HughYauscvi-tools
Trains scVI and scANVI models for batch-corrected single-cell RNA-seq embeddings and differential expression.
HughYauchai1
Predicts protein, nucleic-acid, and small-molecule complex structures with the Chai-1 diffusion model.
HughYauscgpt
Embed and annotate single-cell expression data with the scGPT foundation model.
HughYaudiffdock
Predict small-molecule binding poses against protein structures with DiffDock-L diffusion docking.
HughYauremote-compute-modal
Runs GPU jobs on the user's own Modal account via sandboxed containers.
HughYaumanaged-model-endpoints
Registers local container or remote API model endpoints with a daemon for on-demand inference.
HughYauliterature-review
Find, verify, and synthesize scientific literature using Crossref and OpenAlex APIs.
HughYauskill-creator
Create, evaluate, and iteratively improve AI agent skills with benchmarked testing.
HughYauindication-dossier
Generates a structured therapeutic indication dossier covering epidemiology, biology, standard of care, and clinical trials.
HughYaufair-esm2
Extract protein embeddings, mutation scores, and contact maps using Meta AI's ESM-2 models.
Frequently Asked Questions About Hugh
FAQPage SchemaWhat tasks can I accomplish with HughYau's skill registry?▼
You can predict biomolecular structures with AlphaFold2, ESMFold2, Boltz-2, Chai-1, and OpenFold3; design protein sequences via ProteinMPNN and LigandMPNN; dock small molecules with DiffDock-L; analyze single-cell RNA-seq with scVI and scGPT; and produce publication-grade matplotlib figures and literature reviews.
Who is the target audience for these skills?▼
Computational biologists, structural biologists, protein engineers, and single-cell genomics researchers are the primary users. A secondary audience is software engineers who want disciplined practices like test-driven development, systematic debugging, Git worktrees, and subagent-driven code review.
How do the compute and model-serving skills work in practice?▼
The remote-compute-ssh skill orchestrates job submission and result harvesting on SSH and SLURM hosts, while remote-compute-modal runs GPU jobs in sandboxed Modal containers. The managed-model-endpoints skill registers local container or remote endpoints with a daemon for on-demand inference.
What prerequisites are needed to run the structural biology skills?▼
Structure prediction skills require access to GPU compute, either through SSH/SLURM clusters configured via compute-env-setup or a Modal account. Inputs are standard formats such as PDB backbone structures, DNA or protein sequences, and small-molecule definitions for docking and co-folding models.
Does the registry cover scientific communication and figure preparation?▼
Yes. The scientific-visualization, figure-composer, and figure-style skills produce publication-grade matplotlib, seaborn, and plotly figures, while literature-review synthesizes sources via Crossref and OpenAlex, and paper-narrative restructures manuscript story arcs from a handling editor's perspective.