What does this skill do, and when should you use it?
This is a Claude Skill for academic presentations: conference talks, seminar slides, thesis defenses and grant briefings. It enforces a communication-first content and design standard: every slide title must be a complete sentence stating the takeaway, the titles alone must read as a full argument (the ghost deck test), each results slide carries one exhibit annotated with the key finding, and every borrowed figure or data point gets an in-slide citation plus a closing References slide. It explicitly overrides Claude's default design-forward style — white backgrounds, a single sans-serif font, three colors maximum, no decorative icons. It does not generate the .pptx file itself; it works alongside Anthropic's built-in PPTX skill, which handles technical implementation while this skill governs content, argument structure and design standards.
Once loaded, Claude first identifies the presentation type (defaulting to 'Structured Argument' mode), then produces a slide-by-slide outline (title, action title, exhibit type), confirming with the user before building if the deck exceeds 10 slides or the content is complex. The ghost deck test is run at planning time. Design standards are then applied: white backgrounds on content slides, one sans-serif font throughout, maximum three colors, a type hierarchy (action titles 24–28pt bold, body ≥20pt, citations 12–14pt), left-aligned body text, figure-left/bullets-right on results slides, roughly 40 words of body text maximum per slide, and no decorative icons or full-bleed backgrounds. Finally it follows the PPTX skill's QA procedure plus an academic-specific checklist (action titles, ghost deck test, in-slide citations, References slide, conclusions slide as the last non-appendix slide). Technical implementation depends on markitdown[pptx], pptxgenjs, LibreOffice and Poppler.
- A graduate student preparing a conference talk who wants one convincing claim rather than the whole paper on slides
- A PhD candidate building thesis-defense slides with action titles and a proper References slide
- A researcher preparing a grant briefing that needs structured argument and the key finding highlighted on charts
- A lab-meeting presenter who wants ~40 words per slide so the audience listens instead of reads
- Anyone borrowing figures or data who wants Claude to automatically apply in-slide citations and a References slide
How do you install this skill?
- License metadata contradiction: SKILL.md claims 'Proprietary' and points to LICENSE.txt, but the repo LICENSE is MIT and no LICENSE.txt exists — confirm actual terms with the publisher before use.
- The skill depends on an external PPTX skill for file generation; if that skill is missing or changes behavior, key paths may fail.
- Static review only: no code executed; PptxGenJS snippets and QA procedures are unverified, and example data are fictional.
- Typography rules assume English academic contexts (single Arial/Calibri/Helvetica face, 20pt floor); Chinese fonts and layout require user adjustment.
- No version number, changelog, or maintenance commitment; future update behavior is unpredictable.
- Shell / CLI
- Local filesystem
markitdown[pptx] (pip)pptxgenjs (npm global)LibreOffice (soffice)Poppler (pdftoppm)
- On the GitHub repo page, click Code → Download ZIP; 2. In claude.ai, go to Customize → Skills; 3. Upload the zip file; 4. Confirm the skill appears in your skills list and is toggled on. Prerequisite: enable code execution and file creation in Settings → Capabilities. The repository does not document a manual Claude Code install path (e.g., placing the folder in ~/.claude/skills) or installation on other platforms.
git clone --depth 1 https://github.com/Gabberflast/academic-pptx-skill.git ~/.claude/skills/academic-pptx-skillGenerated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Make slides for my conference paper on X
- Build a deck for my thesis defense
Just ask naturally, e.g., 'Make slides for my conference paper on X' or 'Build a deck for my thesis defense'. Claude detects the academic context and loads the skill automatically. It works alongside Anthropic's built-in PPTX skill: this skill sets content and design standards, while the PPTX skill generates the .pptx file. For decks over 10 slides, Claude presents an outline for confirmation before building.
What are this skill's strengths and limitations?
- Grounded methodology: draws on Barbara Minto's Pyramid Principle, Naegle (2021) 'Ten simple rules' in PLOS Computational Biology, and standard consulting/academic practice
- Explicit QA mechanisms: the ghost deck test and an academic-specific checklist make quality verifiable rather than impressionistic
- Clearly overrides Claude's default flashy style, preventing icon-decorated, gradient-heavy non-academic slides
- The repo ships a PDF summarizing the advice, readable without Claude
- Depends on Anthropic's built-in PPTX skill; it cannot produce files on its own
- Requires a dependency chain: markitdown, pptxgenjs, LibreOffice, Poppler
- No test suite in the repository and no showcased example outputs
- Strict design rules (no icons, three-color cap, ~40-word limit) may feel limiting; the 'Visual/Narrative' branch for science communication offers less concrete guidance
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Academic Presentations Skill (academic-pptx) this page | 52 · Use with care | ★ 1.2k | 3mo ago | MIT |
| deepPPT | 67 · Recommended | ★ 28 | 2mo ago | MIT |
| MBB Decks | 53 · Use with care | ★ 20 | 5mo ago | MIT |
| PPT Master | 64 · Recommended | ★ 62k | 11d ago | MIT |
| PPTX Presentation Skill ✓ Anthropic · Official | 54 · Use with care | ★ 181k | 16d ago | — |
The repository explicitly positions this as a complement to Anthropic's built-in PPTX skill: the latter handles technical .pptx generation and editing, while this skill governs content, argument structure and design standards; they are meant to be used together.
How did FollowSkills review this skill?
Purely instructional content with no script execution, network calls, or sensitive data handling; least privilege is inherently satisfied and dependencies are disclosed. Deductions: no explicit rollback/confirmation mechanisms, behavior depends on an external PPTX skill outside this repo, publisher identity unverified.
Internally consistent, well-structured instructions with a QA checklist and plausible happy path. Static-only review: no tests, no CI, no committed execution evidence; PptxGenJS snippets unverified and contain minor errors (e.g., '²aboratory' typo, hypothetical data); failure feedback for abnormal input is delegated to the external PPTX skill. Capped at 10 per static calibration.
Clear triggers (conference talks, defenses, seminars) and a documented mode split between structured argument and narrative, with default fallback. Deductions: no declared support for Chinese/non-Latin typography (single-font assumptions suit English academic contexts); no mainland-China reachability problem (local installs), but potential trigger conflict with the built-in PPTX skill is not addressed.
Good layered documentation (SKILL.md entry point plus two guide files plus README) and complete install notes. Deductions: SKILL.md declares 'Proprietary. LICENSE.txt has complete terms' while the repo LICENSE is MIT and no LICENSE.txt exists — contradictory license metadata; no versioning, changelog, or stated maintenance/update path.
High methodological value (action titles, SCR structure, exhibit discipline, citation standards) with clear marginal value over generic PPTX generation, and instructions usable directly with Claude's built-in PPTX workflow. But static review cannot verify outputs are directly usable; value claims rest on user execution. Capped at 6.
Cites checkable external sources (Minto, Naegeli 2021 PLOS) and the code snippets are auditable material. But no tests, no CI, no third-party execution evidence; the auto-invocation claim is not statically verifiable, and fictional example data are not clearly flagged as such. Scored 3.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →