agent skill · paper → anything

Turn your paper
into reach.

Turn one academic paper into slides, a poster, a project page, Xiaohongshu & WeChat posts.

50+papers tested
5output formats
1command to install
100%saved locally
A prism splitting one white beam into five colors — one paper becoming five forms

↑ All generated by paper2anything

Capabilities

One paper, five faces

Each subdirectory is a standalone, auto-triggered skill. They share one environment and credentials; workflows, outputs and publishing stay independent.

.pptxA presentation slide title page generated by paper2slides

paper2slides

Pull key points, set the layout, place figures — an end-to-end deck you can present as-is.

→ .pptx"Make slides from this paper"
poster.pngA conference poster generated by paper2poster

paper2poster

Pick figures, set columns, hand-lay it out, then refine with visual and blind-review checks — a poster ready for the conference hall.

→ poster.png + html"Paper to poster"
index.htmlA single-page project homepage generated by paper2html

paper2html

Build a deploy-ready single-page project site — a landing page for GitHub Pages.

→ index.html"Paper to webpage"
XiaohongshuA Xiaohongshu vertical cover generated by paper2xhs

paper2xhs

Frame the angle, write title, body and tags, auto-generate a vertical cover, and post on-account after you confirm.

→ post + cover"Paper to Xiaohongshu"
WeChatA WeChat banner cover generated by paper2wechat

paper2wechat

Outline it, write a deep-dive long read, add figures and a cover, push to the WeChat drafts in one go.

→ long read + cover"Paper to WeChat"
Real output

Really generated, not placeholders

Every image below comes from a full run on a real paper — not cherry-picked, shown as-is.

A collage of slide title pages generated from 50+ papers
paper2slides · Slides

Diverse papers, design at will

No templates, no clones. Each paper is read on its own and picks colors, layout and figures that fit its character.

Auto paletteHD figuresVisual QA
A generated academic conference poster
paper2poster · Poster

A poster ready for the conference hall

Multi-column layout with hand-placed figures and equations, refined via geometry checks and blind review. Exported as high-res PNG and editable HTML.

Multi-columnGeometry checkBlind-review
paper-project.github.io
A generated single-page project homepage
paper2html · Project page

A project page ready to deploy

A self-contained single-page site: inline styles, relative-path figures, responsive layout, passing QA at desktop and mobile widths. Open it and go — push to GitHub Pages to ship.

Single fileResponsiveRender QA
Xiaohongshu vertical cover
paper2xhs + paper2wechat · Chinese social

Title, body, cover — all ready

Turn hardcore papers into Chinese social language: catchy titles, readable copy, plus auto-generated vertical / horizontal covers. Can post straight to Xiaohongshu, WeChat to drafts.

Cover artDirect to XiaohongshuWeChat drafts
Philosophy

Scripts do the grunt work; the agent designs by hand

Every skill is agent-led and coordinated: parsing, figure-cropping and publishing call small tools; understanding the paper, designing the layout and writing the copy are done by the agent, confirming with you at key points.

01
Input

Paper PDF

Give a path, or just say what you want.

02
Mechanical · Script

MinerU parse

Cloud-parsed into structured text and figures, with figures re-cropped in high definition.

03
Core · Agent

Leads design & writing

Reads the paper, sets direction, writes the product by hand, confirms at key points.

04
Mechanical · Script

QA gate

Checks missing figures / broken links / render distortion / content fidelity, fixed to zero per report.

05
Output

Lands beside the paper

The final product is copied next to the PDF — open and use.

No two alike

Same skill, but every paper gets a design built from scratch — palette, structure and layout all differ; the same template is never reused.

Facts only, never fabricated

Numbers, authors and citations come only from the parsed source; gaps are filled by the agent reading the full paper, never invented.

Blind-review QA

Each product is rendered to an image and picked apart by a blind-review agent — missing figures, distortion, overflow, fidelity — looped until zero.

Taste, encoded

Every skill carries its own design language and layout taboos — what to do and what never to do — so outputs avoid generic templates and AI slop.

Quick start

Say what you want, or use a slash command

The matching skill triggers automatically. Either way works:

① Just say what you want
"Make slides from paper.pdf"
"Make a conference poster from this paper"
"Turn this PDF into a project page"
"Post this paper to Xiaohongshu"
"Write this paper up for WeChat"
② Slash command + PDF path
/paper2slides path/to/paper.pdf
/paper2poster path/to/paper.pdf
/paper2html   path/to/paper.pdf
/paper2xhs    path/to/paper.pdf
/paper2wechat path/to/paper.pdf
One-command install (Claude Code)
git clone https://github.com/QuZhan51496/paper2anything.git && cd paper2anything
bash tools/install-linux.sh --create-env --shell-init    # Linux
bash tools/install-macos.sh --create-env --shell-init    # macOS
Acknowledgements

Built on these open-source works