Skip to content

Latest commit

 

History

58 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 

Repository files navigation

arXiv Claude Code Skills

651 curated Claude Code skills derived from state-of-the-art arXiv research papers (Feb 2026). Each skill encodes a research technique as an actionable workflow with step-by-step instructions, concrete examples, and best practices.

Every skill has been individually reviewed for practical usefulness -- skills requiring GPU training, model weight access, or custom hardware have been removed. Only actionable techniques that Claude Code can actually implement remain.

Quick Stats

Metric Value
Total skills 651
Categories 17
Avg lines per skill 211
Avg usefulness rating 5.5/10
Highly rated skills (7+/10) 130
Skills with workflows 99%+
Skills with examples 99%+

How to Use

Install a single skill

Copy any skill folder to your Claude Code skills directory:

# Clone the repo
git clone https://github.com/ndpvt-web/arxiv-claude-skills.git

# Copy a specific skill
cp -r arxiv-claude-skills/skills/mermaid-memory-enhanced-retrieval-reasoning ~/.claude/skills/

Claude Code will automatically load the skill's name and description from the YAML frontmatter and use it when relevant to your queries.

Install multiple skills

# Copy an entire category (find skills via INDEX.md)
for skill in mermaid-memory-enhanced-retrieval-reasoning a2rag-adaptive-agentic-graph compactrag-reducing-calls-token; do
  cp -r arxiv-claude-skills/skills/$skill ~/.claude/skills/
done

Browse by category

See INDEX.md for the full categorized listing with descriptions, ratings, and direct links.

Categories

Each skill is assigned to exactly one category based on Claude analysis. Counts add up to 651.

Category Count Avg Rating What it covers
Evaluation & Benchmarking 92 4.9/10 Benchmarks, metrics, scoring, model assessment
Security & Safety 80 5.9/10 Jailbreaks, adversarial attacks, guardrails, prompt injection, red teaming
Domain-Specific 60 5.1/10 Medical, legal, financial, clinical, robotics applications
Agentic Systems 54 5.7/10 Autonomous agents, tool use, planning, task decomposition
Code & Software Engineering 46 6.6/10 Code generation, bug detection, testing, refactoring, repair
Reasoning & Chain-of-Thought 46 5.7/10 Chain-of-thought, logical inference, step-by-step reasoning, math
Multi-Agent Systems 45 5.3/10 Agent collaboration, swarms, orchestration, debate frameworks
RAG & Retrieval 42 5.6/10 Retrieval-augmented generation, search, reranking, chunking
Efficiency & Optimization 34 5.4/10 Quantization, pruning, compression, acceleration
Data Processing 28 6.0/10 ETL, parsing, extraction, annotation, pipelines
Prompt Engineering 27 5.8/10 In-context learning, few-shot, instruction design, prompt optimization
NLP & Text 24 4.9/10 Classification, summarization, translation, NER, QA
Multimodal 21 5.0/10 Vision-language, audio, video, speech, cross-modal reasoning
Memory & Context 20 5.6/10 Long-context handling, KV cache optimization, context compression
Knowledge Graphs 18 5.2/10 Graph-based knowledge, ontologies, entity relations
Explainability 7 5.0/10 Interpretability, attribution, transparency, causal analysis
Fine-tuning & Training 7 4.3/10 RLHF, GRPO, distillation, curriculum learning, reward models

Skill Structure

Each SKILL.md follows this format:

---
name: "skill-name"
description: "What this skill does and when to trigger it..."
---

# Paper Title

Overview of the technique and when to apply it.

## When to Use
- Specific scenarios...

## Core Technique
The method from the paper, explained for implementation.

## Step-by-Step Workflow
1. Concrete steps...

## Examples
Worked examples with input/output.

## Best Practices
Do's and don'ts.

## Limitations
When NOT to use this technique.

## Reference
Link to the arXiv paper.

How These Were Generated

Skills were generated by a parallel pipeline that:

  1. Scraped 10,000 papers from arXiv (cs.AI, cs.CL, cs.SE, cs.CR, etc.)
  2. Scored and ranked papers by practical applicability (method novelty, implementation signals, code availability)
  3. Selected the top 2,000 papers for skill conversion
  4. Generated skills using Claude CLI (claude --print) with rich prompts containing the paper's full abstract, category, and technique analysis
  5. Filtered out non-viable papers (pure surveys, benchmark-only, position papers)
  6. Quality-checked: 99%+ have structured workflows and concrete examples
  7. Individually reviewed each skill with Claude for practical usefulness, removing 382 impractical skills (those requiring GPU training, model weight access, or custom hardware)

The pipeline ran 4 parallel Claude CLI workers processing papers concurrently, with automatic resume, fallback generation, and progress checkpointing.

Repo Structure

README.md          -- This file
INDEX.md           -- Full categorized skill index with ratings and searchable tables
skills/
  skill-name/
    SKILL.md       -- The skill file (install this to ~/.claude/skills/)

Limitations

  • Skills encode methodology from papers, not trained models. Results depend on Claude's general capabilities.
  • Papers were evaluated on specific benchmarks; real-world tasks may differ.
  • Skills reference techniques current as of February 2026. APIs and libraries may change.
  • Loading many skills simultaneously increases context usage. Install selectively based on your domain.

License

Skills are derived from publicly available arXiv paper abstracts and metadata. Individual papers retain their original licenses. The skill generation pipeline and templates are provided as-is.


Generated: February 2026 | Papers source: arXiv | Generator: Claude CLI parallel pipeline


Keywords

For the search engines and the curious humans who find things by typing random words

arXiv skills, research paper to skill, AI research tools, Claude Code skills library, academic AI tools, paper implementation, research automation, AI skill generator, arxiv paper tools, scientific AI skills, research to code, Claude Code extensions, AI agent skills collection

About

Claude Code Skills generated from arXiv research papers - AI-powered skill library

Resources

Stars

14 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors