Guide · Agent skill
How to use the Paper Poster (HTML) Generator agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The paper-poster-html skill teaches an AI agent to build an academic conference poster from a research paper, for venues such as ICML, NeurIPS, ICLR or CVPR. This guide explains what the skill makes the agent do, how to install it, how to trigger it with plain requests, and what it needs to run. By the end you will know how to set it up and what to check before you print.
Paper Poster (HTML) Generatorwanshuiyin/Auto-claude-code-research-in-sleep · MIT Download the skill Quick start
- 1Install the base package. Install the skills/skills-codex package first, as the source requires for this overlay.
- 2Install paper-poster-html. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills.
- 3Configure the gemini-review bridge. Set up the gemini-review MCP bridge. Without it the skill stops rather than reviewing its own work.
- 4Ask for a poster. Name your paper directory or PDF and the venue, for example ICLR, and use a trigger phrase such as "conference poster".
- 5Let the agent iterate on gates. The agent writes the HTML, measures the layout against its hard gates, and fixes failures through the gate process.
- 6Review the output. Read the gate report and claim evidence files, then check the rendered PDF yourself before printing.
What the paper-poster-html skill makes the agent do
The skill writes the poster as one HTML and CSS file sized to the exact print canvas, using an @page size rule. The agent then renders that file to PDF through Playwright print emulation in headless Chromium.
Layout is checked by measurement rather than by eye. The source says the screen preview is misleading and that only print emulation at the correct viewport shows the real result. The poster uses real figures from the paper and a two-hue design-token system.
- Output is a single HTML/CSS file plus a print-ready PDF.
- Layout is checked with measurement-driven hard gates.
- Figures come from the paper rather than being redrawn.
- A second model, Gemini, audits claims against the paper and reviews the final rendered image.
@page { size: W H }How to install the paper-poster-html skill
The skill is distributed as a ZIP from the download link on this page. Unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill limited to one project.
You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
This skill is an overlay. The source says to install it after the skills/skills-codex/* package, because it is meant to replace the base reviewer with Gemini.
- Install the base skills/skills-codex package first.
- Then install paper-poster-html, either to ~/.claude/skills/ or to a project's .claude/skills/.
- Or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
How to trigger the skill with natural requests
You do not run the skill with a command. The agent loads it when your request matches its description. The source lists these trigger phrases: "做海报", "poster", "conference poster", "paper poster", and requests to design or redo a research poster.
Give the agent your paper directory or PDF, and name the venue and canvas if you have them. The argument hint in the source shows options such as venue, canvas size, landscape orientation and venue colors.
- "Make a NeurIPS conference poster from paper/ with a 185x90cm landscape canvas."
- "做海报 using the PDF at drafts/main.pdf, venue ICLR."
- "Redo my research poster for CVPR using the figures in the paper folder."
What the skill needs to run
The skill needs Claude Code or a compatible SKILL.md agent, plus headless Chromium or Playwright for print emulation and PDF output. It also needs the gemini-review MCP bridge configured, because both cross-model review steps call Gemini through it.
If the gemini-review bridge is unavailable, the skill stops and tells you what to configure. It does not fall back to self-review. The skill runs scripts and uses the network.
- Claude Code or a compatible SKILL.md agent.
- Headless Chromium or Playwright.
- The gemini-review MCP bridge configured.
- The skills/skills-codex package installed first.
- Scripts run and network access is used.
mcp__gemini-review__review_startSafety notes: scripts and network
The skill's allowed tools include Bash, file read, write and edit, and search. It runs scripts and uses the network, so review what it runs before you approve it in a shared or sensitive project.
Review calls send content to Gemini. The source says the bridge cannot read your local files, so the agent pastes the relevant poster text and paper sections into the prompt and attaches rendered images. Check that you are comfortable sharing that content before you run it.
The license is MIT. The source says core gate machinery is adapted from posterly, which is also MIT.
- Assume paper excerpts and poster text leave your machine during review calls.
- Keep the gemini-review bridge configured only with services you trust.
- Read the gate report and claim evidence file the skill writes before you print.
Tips for better posters
Start from a finished paper draft or PDF, since the claim audit checks every number, equation and attribution against the paper text. Fix every non-OK row in the claim evidence report, or record it as a tradeoff you accept.
Expect several rounds. The source says fixes go back through the gates, not straight to re-review. Check headline numbers, banner readability from 2 m, and that real figures sit inside their cards.
- Provide the paper text so claim checks have a source.
- Read CLAIM_EVIDENCE.md for every non-OK row.
- Check the final PDF yourself before submitting.
What people use it for
- Turning a conference paper draft into a poster for ICML, NeurIPS, ICLR or CVPR.
- Redoing an existing research poster with real figures from the paper.
- Producing a print-ready PDF for a student poster session on a set canvas size.
Frequently asked questions
Is the paper-poster-html skill free?
The source gives the license as MIT. It does not state pricing for the skill, and it depends on tools such as Chromium and a Gemini bridge that you set up yourself.
What do I need to install paper-poster-html?
Claude Code or a compatible SKILL.md agent, the skills/skills-codex package installed first, and the gemini-review MCP bridge configured. You also need headless Chromium or Playwright for PDF output.
Where do I put the skill folder?
Unzip the downloaded ZIP into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. You can also upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
Why does the skill stop instead of reviewing its own poster?
The source says that if the gemini-review bridge is unavailable, the skill stops and tells you what to configure. It does not silently fall back to self-review, because the cross-model review is the point of the check.
Can I use it without Gemini?
Not as written. The source says the two cross-model review calls go to Gemini through the gemini-review bridge. The base skills/skills-codex version uses a GPT reviewer agent instead.
Can I submit the poster without checking it?
The source warns against final submissions without human checking, because the cross-model audit can still miss errors or overclaims. Review the output yourself before you submit.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a researcher or student who already runs the ARIS skill pack with a Gemini review bridge and needs a conference poster built from a paper. Pick a different tool if you want a slide deck, a non-academic flyer, or a setup without the gemini-review bridge.
Download the skill