latex-paper-skills
Eight skills for LaTeX research papers, from topic to compiled PDF, with verified BibTeX citations and gated steps
- 272 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Scripts run LaTeX compilation and write paper files, the task table and figures into the project directory
- The skill goes online for arXiv metadata and citation checks, and the optional bridges call installed Gemini CLI and Claude Code
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/latex-paper-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Scripts run LaTeX compilation and write paper files, the task table and figures into the project directory
- The skill goes online for arXiv metadata and citation checks, and the optional bridges call installed Gemini CLI and Claude Code
This is third-party code. Review the repository files before installing.
What it does
latex-paper-skills installs as a single package and walks an agent through writing a machine-learning or AI paper. The `paper-from-zero` skill takes a topic, searches the literature, frames the contribution and builds an evidence matrix before handing off to a writer. `arxiv-paper-writer` writes surveys in the IEEEtran template and keeps a task table in CSV, while `empirical-paper-writer` adds an experiment plan and result statuses: planned, placeholder, verified. The other skills cover separate steps. `results-backfill` puts real results into a finished draft and builds figures. `latex-rhythm-refiner` adjusts sentence rhythm and leaves every `\cite{}` position alone. Three helper skills cross-check the work through Gemini CLI and Claude Code, and `check-collaborators` confirms those CLIs are installed and responding. The rules are strict. No prose is written until the plan is approved. A task is marked done only after its acceptance criteria pass. Every citation is checked against an online source before it enters `ref.bib`, and inventing citations, results or significance claims is forbidden. Python scripts compile the PDF with `latexmk` or `pdflatex` plus `bibtex`, validate the task table and rank source quality.
Who it is for. Graduate students and researchers who write AI surveys and experimental papers in LaTeX and want the agent to stop inventing citations and results.
Good fit when
- You only have a topic and need a path from literature search and a contribution map to a survey draft in IEEEtran
- Experiments run separately and the draft holds placeholders that must be replaced with real numbers from a CSV
- The draft reads monotonously and sentence rhythm needs evening out while `\cite{}` calls stay where they are
- Before submission you need to confirm every BibTeX entry matches a real source
Not a fit when
- The paper is outside machine learning and AI: the templates and rules target that field
- The experiments must run in the cloud: the skill plans for the budget but does not submit jobs
- There is no LaTeX environment or Python 3.10: without `latexmk` or `pdflatex` the PDF will not build
- You want a quick draft without agreeing a plan: the workflow forbids writing prose before approval
Example request
Use latex-paper-skills. My topic is speeding up large language model inference. Start with a literature search and a contribution mapLimitations
- The author tested the showcase projects on GPT-5.2 xhigh and expects better results from GPT-5.4. Test coverage is limited, so bugs are possible.
- Installation needs Node.js 22.20 or newer, the scripts need Python 3.10 or newer, and the PDF needs a LaTeX environment.
- Collaboration with Gemini and Claude needs installed and authenticated CLIs.
- The skill does not launch cloud jobs: real results are produced separately and placed in `paper/results/`.
- The package version is 0.1.0 and the ClawHub listing is 1.0.2: they are separate numbers.
How to disable. Delete the `latex-paper-skills` folder from the agent's skills directory where the installer placed it. A command with `--global` installs into the user directory, without it the skill sits in the project. Your paper files stay in the project folder.
FAQ
How do I install latex-paper-skills in Claude Code?
From the project directory run `npx latex-paper-skills@latest --agent claude-code --copy --yes`, which needs Node.js 22.20 or newer. Then tell the agent: "Use latex-paper-skills. My topic is: your topic".
How do the survey and empirical writers differ?
`arxiv-paper-writer` handles surveys and citation checks, and `empirical-paper-writer` adds an experiment matrix, baselines and result statuses. Until the numbers exist they are marked as placeholders.
Does the agent invent citations and results?
The rules forbid it. A citation enters `ref.bib` after a check against an online source, and results come only from real files in `paper/results/`.
What is needed to build the PDF?
A LaTeX environment: `latexmk` or `pdflatex` together with `bibtex`, and Python 3.10 or newer for the scripts. The installer needs Node.js 22.20 or newer.
Can I install individual skills from the bundle?
No, install the whole package. The separate folders under `.codex/skills/` lack the shared dependencies from `_shared`, so the skills would not work without them.
Discussion
No questions or reviews yet.
Related
SkillsEight skills for finished papers and theses in LaTeX and Typst: compile, references, language polish, cover letters and a reviewer-style audit
Takes an agent from a topic to an arXiv review paper in LaTeX IEEEtran: approved plan, every citation checked against a live source, PDF build
A Claude Code skill for research papers: a five-stage pipeline, section rules, a style audit and an independent red-team review.
The paper-spine skill takes a topic, materials and data through a scientific paper: literature, text, figures, review and Word, LaTeX and PDF files
Skill with 30 prompt templates for research papers in English and Chinese, from literature review to grant proposals
A Claude Code skill: describe a document in plain words and get a compiled LaTeX PDF. Templates, build scripts and OCR of PDFs and handwritten notes
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| latex-paper-skills (this card) | no | 272 | |
| paper-writing-skill | no | 250 | a single skill for paper prose, without the task-table pipeline and citation checks |
| Qinyan Academic Skills | no | 1,011 | academic skills without mandatory plan approval gates |
README in short
The README, in English and Chinese, describes a bundle of eight skills with a pipeline from topic to PDF and optional multi-agent collaboration. It has skill tables, a stage diagram, a quick start via npm and the Skills CLI, prompt examples and two finished showcase projects: one empirical and one survey.
README badge
Original SKILL.md
--- name: latex-paper-skills description: > Write, revise, and validate ML/AI research papers in LaTeX. Use for topic-to-paper planning, review papers, empirical papers, verified result backfill, citation audits, and academic prose refinement. license: MIT metadata: version: "0.1.0" source: https://github.com/yunshenwuchuxun/latex-paper-skills runtime: Python 3.10+ and web research; LaTeX for PDF compilation; optional Gemini and Claude CLIs. --- # LaTeX Paper Skills This package includes eight workflows, their shared Python tools, and IEEE LaTeX templates. Select the workflow that matches the user's deliverable and read its instructions before starting. Read further references only as that workflow requires them. | Request | Workflow to read | | --- | --- | | A research topic needs contribution framing, evidence planning, and a paper route | [paper-from-zero](references/skills/paper-from-zero/workflow.md) | | Write or revise a review or survey; validate citations in an existing LaTeX project | [arxiv-paper-writer](references/skills/arxiv-paper-writer/workflow.md) | | Plan or draft an empirical paper, including experiment contracts and placeholders | [empirical-paper-writer](references/skills/empirical-paper-writer/workflow.md) | | Fill an empirical draft with real, verified experiment results | [results-backfill](references/skills/results-backfill/workflow.md) | | Improve sentence and paragraph rhythm while preserving meaning and citations | [latex-rhythm-refiner](references/skills/latex-rhythm-refiner/workflow.md) | | Check optional Gemini and Claude collaborators | [check-collaborators](references/skills/check-collaborators/workflow.md) | | Expand candidate directions or literature using Gemini | [collaborating-with-gemini](references/skills/collaborating-with-gemini/workflow.md) | | Audit claims and evidence using Claude | [collaborating-with-claude](references/skills/collaborating-with-claude/workflow.md) | ## Resolve installed resources Determine the absolute directory containing this `SKILL.md`; call it `<package-root>`. The bundled skill tree is `<package-root>/references/skills/`. All eight workflows and `_shared/` are included there; no separate sibling-skill installation is needed. When following a bundled workflow: - Resolve `scripts/`, `assets/`, and `references/` relative to that workflow's directory, rather than the user's current working directory. - Resolve `../<other-skill>/` relative to the same directory. A reference to another workflow means read its bundled `workflow.md`. - Replace legacy `.codex/skills/` command prefixes with the absolute bundled