Skip to main content
affpapa.org
RU EN
skill CX Codex

paper-to-code — Claude Code skill: what it does, how to install

Turns an ML research paper into a runnable code repository. Useful when reproducing published experiments and models without manual scaffolding.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add lingzhi227/agent-research-skills --skill paper-to-code -a claude-code
Open on GitHub

What paper-to-code does

The skill implements the Paper2Code pipeline to reconstruct ML projects through planning, per-file analysis, and dependency-ordered coding. During the planning turn, the agent extracts methodology and training details into config.yaml, then sets up Mermaid diagrams. These diagrams act as strict interface contracts across every subsequent step.

Generation follows a strict order, starting with dataset_loader.py and proceeding to model.py, trainer.py, evaluation.py, and main.py. It produces a complete implementation with zero placeholders or fabricated hyperparameter values. If execution fails, an iterative debugging step applies SEARCH/REPLACE diffs until the code runs cleanly.

It requires a PDF path, raw paper text, or URL as input. Any ambiguity in the published methodology is flagged explicitly instead of being masked by improvised code fixes.

What it helps with

  • extract architecture and hyperparameters from an ML paper
  • generate class and sequence diagrams from research text
  • build a complete runnable repository in dependency order
  • fix runtime errors using minimal diff replacements

Example prompts for the agent

  • Reproduce the methodology from paper.pdf into a runnable project
  • Extract the model architecture and training configuration from paper.txt
  • Generate reproduced code and a reproduce.sh script from the research paper

How to install paper-to-code

1Skills CLI (Vercel)

npx skills add lingzhi227/agent-research-skills --skill paper-to-code -a claude-code

2Manually: put the files in place

~/.claude/skills/paper-to-code/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/paper-to-code/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about paper-to-code

How to install paper-to-code in Claude Code?

Run npx skills add lingzhi227/agent-research-skills --skill paper-to-code -a claude-code in your terminal.

What files does paper-to-code generate?

It builds a reproduced_code directory containing config.yaml, dataset_loader.py, model.py, trainer.py, evaluation.py, main.py, requirements.txt, and reproduce.sh.

How does paper-to-code handle missing hyperparameter details?

The workflow forbids fabricating configuration values. If details are ambiguous in the text, the skill explicitly notes them instead of guessing.

Similar skills

Everything for AI and ML →
research-paper-writing Orchestra Research Guides the end-to-end ML research paper workflow from experiment design to LaTeX drafts for top AI conferences. CX Codex +1 493installs implement-paper marimo-team Walks through a research paper with you and builds a focused interactive marimo notebook from the part you care about. CX Codex +1 2,700installs huggingface-paper-publisher huggingface Indexes arXiv papers on Hugging Face Hub, links them to model and dataset cards, verifies authorship and builds markdown drafts from templates. CX Codex +1 1,700installs daily-paper-generator Galaxy-Dawn A skill that searches arXiv and bioRxiv for recent papers on a topic and writes bilingual reviews of the ones it picks. CX Codex +1 410installs shap K-Dense-AI Guides AI agents in explaining and auditing machine learning predictions with SHAP. Useful for selecting explainers, checking additivity, and rendering attribution plots. CX Codex +1 2,000installs pytorch-lightning K-Dense-AI Guides agents through structuring PyTorch Lightning 2.6.6 code, from LightningModules to distributed trainers. Useful when scaling training across GPUs. CX Codex +1 2,000installs

From the same set30

algorithm-design Turn a method description or implementation into LaTeX pseudocode and Mermaid diagrams. Use it to formalize an algorithm and check its representation against code. CX Codex +1 1,700installs survey-generation Walks an agent through paper collection, outlining, writing and citation checking, then hands back a LaTeX survey with a .bib file. For when you need an actual literature review. CX Codex +1 1,700installs symbolic-equation Discovers scientific equations from numerical datasets using LLM-guided evolutionary search. Useful for symbolic regression and extracting interpretable physical laws. CX Codex +1 1,600installs experiment-code Writes and iterates on ML experiment code for a research paper: training, evaluation, debugging, plots. Useful when you need a batch of runs and a comparison. CX Codex +1 1,600installs atomic-decomposition Splits a research idea into atomic concepts, then finds the math formula in papers and the matching code for each one. CX Codex +1 1,600installs paper-compilation Compiles a LaTeX paper to PDF, catches build errors and checks citations and references. Useful when pdflatex throws errors and the log is a mess. CX Codex +1 1,600installs slide-generation Turns a completed LaTeX paper into Beamer presentation slides or a poster. It helps researchers prepare talks and poster sessions directly from paper source files. CX Codex +1 1,600installs self-review This skill reviews an academic paper with the NeurIPS form, three reviewer roles, and score aggregation. Use it to assess a draft before submission. CX Codex +1 1,600installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor