Scientific Slides

A scientific presentation skill with PowerPoint, PDF, and LaTeX Beamer workflows, templates, and visual review.

Scientific Slides is an independent open-source skill from davila7's official repository. The verified path is cli-tool/components/skills/scientific/scientific-slides, and it contains the SKILL.md file, detailed references, three LaTeX Beamer templates, and executable Python scripts. According to the provider, it targets scientific conference talks, seminars, research presentations, journal clubs, teaching talks, grant pitches, and dissertation or thesis defenses. It is not a presentation editor and not a hosting service. Its product boundary is a reusable set of instructions with templates and local review tools that a compatible coding agent can apply to explicitly selected project files.

Structure and planning

The source starts with deck planning: each slide should have a title, a central message, visual elements, and appropriate citations. It recommends a clear narrative rather than a sequence of data dumps and asks the author to consider figures, references, and research context visibly. Its guidance distinguishes talk types and time budgets. It names roughly one slide per minute as a general rule, longer treatment for complex result slides, separate phases for introduction, methods, results, discussion, and conclusion, and timed rehearsals. These are provider-authored editorial aids, not a replacement for conference instructions or subject-matter review.

PowerPoint, PDF, and Beamer

For PowerPoint, the skill describes a separate PPTX workflow: the included generator can create visual assets first, after which text and images are placed into an editable deck through the PPTX skill or template editing. The component includes a PowerPoint design guide, but it does not contain a PowerPoint library or guarantee that an external PPTX skill directory is installed. The PDF workflow creates individual image slides and combines them with slides_to_pdf.py. For LaTeX Beamer, the component provides templates for an approximately fifteen-minute conference talk, a longer seminar, and a defense. These Beamer files are editable starting points for equations, code, algorithms, references, and institutional adaptation; they are not finished claims about any research topic.

Rendering and validation

The pdf_to_images.py and validate_presentation.py scripts support repeatable technical checks. Depending on the input, validation analyses PDF, PPTX, or LaTeX files and checks items such as slide count, dimensions, file size, text size, and possible LaTeX compilation with pdflatex. The visual_review_workflow.md reference then requires conversion to images and human inspection for overflow, overlaps, readability, contrast, figure size, and narrative logic. Pillow, PyMuPDF, PyPDF2, python-pptx, and in some paths NumPy are expected Python dependencies; their versions are not pinned in this repository component. Beamer validation requires a local TeX installation.

Data, files, and external models

Image generation is optional and not locally autonomous: generate_slide_image.py calls the OpenRouter image and review model chain configured by the source and requires an OPENROUTER_API_KEY. Prompts and attached figures can therefore reach an external model provider. Use only approved, minimized, non-confidential inputs. Before every run, verify input paths, output paths, image rights, fonts, quotations, research data, and generated PDFs. Tokens, passwords, private keys, and raw personal data must not be placed in prompts or decks. Generated slides are drafts: facts, sources, statistical interpretation, accessibility, rights, rehearsal timing, and final presentation require responsible human approval.

Boundary against other presentation and diagram skills

Scientific Slides overlaps with general presentation and diagram tools but does not replace them. Frontend Slides in the catalog targets animated HTML presentations and their web or PPTX conversion. Mermaid Diagrams and Excalidraw Diagram Generator focus on text or drawing-oriented diagrams, not scientific deck structure, Beamer templates, timing, or presentation validation. This skill is consequently bounded as a scientific planning, template, and review workflow, while actual PowerPoint authoring, TeX compilation, and subject-matter approval remain responsibilities of the installed tools and accountable people.

Installation instructions

What the Scientific Slides Skill Offers

Scientific Slides is a skill from the community repository claude-code-templates, which provides a collection of extensions, components, and skills for use with Claude Code. According to the provider, this skill assists in building slide decks and presentations for research talks, for example conference presentations, seminar talks, thesis defense slides, or other scientific talks. It supplies structural and design templates, guidance on timing for individual sections, and visual validation of the result, and it works with both classic PowerPoint and LaTeX Beamer, a typesetting system for presentations widely used in academia.

The Skill's Underlying Philosophy

According to the description stored in the skill, its design follows a clear philosophy: scientific presentations should be visually engaging and research-backed rather than dry, text-heavy slides. In practice this means using high-quality figures, images, and diagrams instead of pure bullet points, incorporating proper citations to establish credibility, keeping text deliberately minimal so that bullet points serve only as verbal prompts rather than carrying the full explanation, and applying a professional, modern design with a strong visual hierarchy and generous white space. The skill also emphasizes a story-driven structure with a clear narrative arc rather than a mere sequence of data dumps.

Step 1: The Suggested Creation Process

For the default workflow, output as PDF slides, the skill describes a multi-step process: first, a detailed plan is created for each slide, capturing the title, key points, and planned visual elements. According to the skill description, an AI image-generation tool called Nano Banana Pro is then used to produce each slide as a complete image. In the final step, the generated slide images are assembled into a single PDF document. This approach differs noticeably from classic slide design in PowerPoint, because entire slide images are generated with AI support rather than individually placed text boxes and shapes.

Step 2: Prerequisites for Use

To use the skill meaningfully, a Claude Code environment with access to read, write, edit, and bash tools is needed, as specified in the permission list stored within the skill. Image generation additionally requires access to the mentioned Nano Banana Pro tool, which is not available by default in every environment. Anyone who instead prefers classic, editable slides in PowerPoint or LaTeX Beamer needs to use the skill's corresponding alternative workflow steps, which rely less on image generation and more on structured text and layout templates.

Practical Value for Researchers

The practical value lies mainly in the fact that researchers, who rarely have time for elaborate slide design, can reach a visually convincing result quickly, one grounded in proven principles of presentation didactics. Instead of spending hours building slide layouts by hand, the skill can serve as a starting point that already accounts for structure, timing, and visual design. This can considerably ease the entry for early-career researchers who still have limited experience with professional presentation design.

Limits and Important Notes

It matters that the Nano Banana Pro-based PDF generation produces fixed, no longer directly editable slide images; later text changes therefore require either regenerating the affected image or switching to the PowerPoint or LaTeX Beamer variant with editable elements. The factual accuracy and scientific rigor of inserted citations and data statements also remain the researchers' own responsibility, since the skill assists with design and structure but does not replace a subject-matter review of the presented research findings.

Free
Provider
davila7
License
MIT
Last reviewed
10.09.2026

Repository and documentation

Categories

Compatible with

Claude Code