Haoke Han
Community@clay-hhk · QingDao
I am Clay, a student in QingDao University. I want to study in Github and try to make some contributions to the community.
Agent Skills by Haoke Han
Showing 183 vetted skills indexed across 2 GitHub repositories.
Clay-HHKplanning-with-files
Creates persistent markdown files for task planning, progress tracking, and knowledge storage.
Clay-HHKui-ux-pro-max
Generates UI/UX design systems with styles, palettes, typography, and stack guidelines via a searchable CLI database.
Clay-HHKCommand Development
Create Claude Code slash commands with frontmatter, arguments, and bash execution.
Clay-HHKnetworkx
Create, manipulate, and analyze graph data in Python.
Clay-HHKscientific-brainstorming
Facilitate scientific ideation and hypothesis generation through structured brainstorming workflows.
Clay-HHKdaily-paper-generator
Automates arXiv search, quality review, and bilingual summarization of recent EEG decoding papers.
Clay-HHKwriting-anti-ai
Detect and rewrite AI writing patterns in English and Chinese prose.
Clay-HHKshap
Attribute model predictions to input features using SHAP values.
Clay-HHKSkill Development
Guides creation of Claude Code plugin skills with progressive disclosure structure.
Clay-HHKumap-learn
Reduce high-dimensional data to low-dimensional embeddings for visualization and clustering.
Clay-HHKtheme-factory
Apply curated color palettes and font pairings to slides, documents, and HTML artifacts.
Clay-HHKsympy
Perform exact symbolic mathematics on algebraic, calculus, and differential equation expressions.
Clay-HHKpaper-self-review
Reviews academic papers against a systematic quality checklist before submission.
Clay-HHKskill-quality-reviewer
Evaluates Claude Skills across description, organization, style, and structure to generate scored quality reports.
Clay-HHKdask
Scale pandas and NumPy workloads across cores or clusters with lazy evaluation.
Clay-HHKexploratory-data-analysis
Detect scientific file types and generate markdown EDA reports with quality metrics.
Clay-HHKdoc-coauthoring
Guides collaborative document creation through context gathering, iterative drafting, and reader testing.
Clay-HHKcitation-verification
Verify academic citations using WebSearch and Google Scholar during paper writing.
Clay-HHKscientific-schematics
Generate publication-quality scientific diagrams from natural-language descriptions.
Clay-HHKpyzotero
Manage Zotero libraries programmatically via the Zotero Web API v3.
Clay-HHKseaborn
Create statistical visualizations from tabular datasets with pandas and matplotlib.
Clay-HHKskill-improver
Applies improvement plans from quality reviews to update Claude Skill files.
Clay-HHKarchitecture-design
Implements Factory and Registry patterns for new registrable components in ML projects.
Clay-HHKgit-workflow
Enforces Git workflow standards covering commits, branching, merging, and conflict resolution.
Frequently Asked Questions About Haoke Han
FAQPage SchemaWhat tasks can I accomplish with Clay-HHK's skill collection?▼
You can fine-tune LLMs with LoRA/QLoRA, quantize models to 4-bit via GPTQ/AWQ/GGUF, serve inference with vLLM or TensorRT-LLM, build RAG pipelines with LangChain/LlamaIndex and vector databases, run mechanistic interpretability experiments, and draft ML conference papers with verified citations.
Who is the target audience for these skills?▼
ML engineers, LLM researchers, and graduate students working on model training, alignment, and deployment. The collection also serves academic writers preparing NeurIPS, ICML, ICLR, or ACL submissions, plus developers building RAG applications and Claude Code plugins.
How do I install and run these skills?▼
Each skill ships as a SKILL.md file with YAML frontmatter declaring dependencies. Install listed packages (e.g., vllm, transformers, peft, deepspeed) into your Python environment, then place the skill folder in your Claude Code plugin directory for automatic discovery and triggering.
Are these skills free and open source?▼
Most skills authored by Orchestra Research carry the MIT license, permitting free use and modification. A subset—including pdf, docx, xlsx, pptx, and frontend-design—are proprietary with terms defined in bundled LICENSE.txt files, so check each skill's license field before redistribution.
What prerequisites do the LLM training skills require?▼
Training skills require PyTorch 2.0+, transformers, and often CUDA-capable GPUs; distributed setups need DeepSpeed, Ray, or FSDP2. Quantization and serving skills need libraries like auto-gptq, autoawq, bitsandbytes, or llama-cpp-python, with some supporting CPU and Apple Silicon inference.