Creating academic presentations is usually a manual, error-prone process. This repo flips that script by treating large language models (LLMs) as a full-fledged compiler for Beamer LaTeX presentations. It wraps the entire lifecycle — from collaborative lecture creation through multi-pass compilation and rigorous quality audits — into a Claude Code skill that enforces strict design and pedagogical standards.
what the beamer skill does and its architecture
At its core, this project is a Claude Code skill, implemented as a large instruction file (~55KB), that transforms prompts into fully compiled Beamer presentations. The skill orchestrates a multi-phase workflow that starts with lecture content generation and goes through phase gates including proofreading, visual auditing, pedagogical validation, and final excellence scoring.
The architecture involves chaining multiple AI coding assistants such as Claude Code itself, OpenAI Codex CLI, Google Antigravity, and VS Code Copilot variants (Copilot Chat, Cline, Cursor). These tools consume the skill’s instructions which encode not only content generation logic but also design constraints and validation rules.
The output is a production-ready Beamer PDF document with no overlays, controlled content density (both upper and lower bounds), and strict WCAG AA color contrast compliance. It also enforces mathematical precision in TikZ diagrams, including audits detecting box overflows visually and semantically. Figures can be extracted from paper PDFs and integrated seamlessly.
The skill supports timing heuristics for talks from brief 5-minute bursts to full 90-minute lectures, adapting slide count and pacing accordingly.
multi-pass automated quality control and design constraints
What distinguishes this skill is its rigorous automated quality control that spans a dozen dimensions humans often miss. The skill encodes a deductive scoring rubric that penalizes violations such as:
- Overlay usage (completely disallowed to maintain clarity)
- Content density outside prescribed bounds, ensuring slides neither cram nor underutilize space
- Color contrast failures against WCAG AA standards, improving accessibility
- Mathematical inaccuracies or visual glitches in TikZ diagrams detected by a visual audit
- Pedagogical validation against 13 formal instructional patterns, ensuring the presentation is not just technically correct but educationally sound
The skill’s multi-phase pipeline involves multiple passes of compilation and review. Each phase acts as a gate, requiring successful checks before proceeding. This design mimics a compiler’s strict type-checking or linting passes but applies them to presentation design and pedagogy.
The codebase is surprisingly compact given the complexity — a single, well-structured instruction file that encodes these rules and workflows. The tradeoff is a heavy reliance on the AI assistant’s ability to interpret and execute these instructions correctly, which depends on the quality and version of the underlying LLM.
quick start
prerequisites
TeX distribution
A full TeX distribution with XeLaTeX is required:
## installation
Clone the repo first:
```bash
git clone https://github.com/Noi1r/beamer-skill.git
Claude Code
Copy the skill directory into your Claude Code skills folder:
mkdir -p ~/.claude/skills
cp -r beamer-skill/beamer ~/.claude/skills/
Restart Claude Code. The skill will be automatically detected and triggered when you mention beamer, slides, lecture, tikz, or related keywords.
OpenAI Codex CLI
Copy AGENTS.md and references/ into your project root:
cp beamer-skill/beamer/AGENTS.md your-project/AGENTS.md
cp -r beamer-skill/beamer/references your-project/references
Codex CLI automatically reads AGENTS.md from the working directory. The main file contains core rules and action summaries; detailed workflows are in references/ and referenced as needed.
Google Antigravity
Antigravity is compatible with the SKILL.md format. Copy the skill directory:
mkdir -p ~/.claude/skills
cp -r beamer-skill/beamer ~/.claude/skills/
The same SKILL.md used by Claude Code works with Antigravity without modification.
VS Code — GitHub Copilot
Copy the Copilot instructions file into your project:
mkdir -p your-project/.github
cp beamer-skill/.github/copilot-instructions.md your-project/.github/
Copilot Chat automatically reads .github/copilot-instructions.md and applies the rules during conversations.
VS Code — Cline
Copy the Cline rules file into your project:
mkdir -p your-project/.clinerules
cp beamer-skill/.clinerules/beamer.md your-project/.clinerules/
Cline automatically loads all files in .clinerules/ as custom instructions.
VS Code — Cursor
Copy the Cursor rules file into your project:
mkdir -p your-project/.cursor/rules
cp beamer-skill/.cursor/rules/beamer.mdc your-project/.cursor/rules/
Cursor automatically loads .mdc files from .cursor.
verdict
This skill is a solid reference implementation for anyone looking to automate the creation of Beamer presentations with strong quality guarantees enforced by AI. It’s especially relevant for educators, researchers, or academic speakers who want to streamline slide deck production without sacrificing design or pedagogical rigor.
The tradeoff is the complexity and size of the instruction file and the dependency on the AI assistant’s correct interpretation. It’s not a drop-in solution for casual users but a specialized tool for those who want automated, multi-dimensional slide quality control.
If you frequently prepare academic talks or lectures and want a way to enforce strict accessibility, design, and pedagogical standards automatically, this repo offers a unique approach worth exploring.
Related Articles
- SkillUI: Automating design system extraction into Claude Code skills — SkillUI reverse-engineers design systems from URLs or repos into Claude Code skill packages using static and visual anal
- ordinary-claude-skills: an extensive local-first library of Claude prompt packages for specialized AI agents — Discover ordinary-claude-skills, a local-first collection of 600+ prompt packages that specialize Claude AI with domain
- daymade/claude-code-skills: a production-hardened plugin marketplace for Claude Code skills — daymade/claude-code-skills offers a robust plugin marketplace with 51 pre-built Claude Code skills and a hardened skill-
- Understanding LLM internals: a hands-on guide to transformers and attention math — A curated repo breaking down large language model internals with numeric attention math, tokenization, and transformer a
- Inside llm-madness: a lightweight GPT transformer training pipeline with built-in visualization — llm-madness offers a Python-built GPT-style transformer training pipeline with tokenizer training, memory-mapped dataset
→ GitHub Repo: Noi1r/beamer-skill ⭐ 227 · TeX