Present Paper

Present-Paper Skill

Purpose

Prepare a polished academic presentation from a research paper. The skill walks through a 5-phase pipeline: paper analysis, supporting research, script writing, slide note injection, and Q&A preparation.

Use it when:

• preparing a journal club or seminar presentation
• presenting a paper for a graduate course
• preparing grand rounds or conference talks based on a published paper
• building speaker notes for an existing slide deck

---

Communication Rules

• Communicate with the user in their preferred language.
• Use English for medical, statistical, and methodological terminology.
• Add pronunciation guides for drug names and technical abbreviations in the user's language.
• Be direct about paper limitations, but frame them constructively.

---

Phase 0: Init & Outline

Step 0a — Load design references (read before drafting outline)

Before collecting inputs, the skill loads three reference files:

1. `references/slide_design_principles.md` — Reynolds (Presentation Zen) + Duarte (Slide:ology Glance Test™) + Knaflic (Storytelling with Data preattentive attributes) + Tufte (Cognitive Style of PowerPoint). Defines the 5 design principles, reading-time budgets per audience, cognitive-load ceilings, and the anti-patterns this skill is built to avoid. Read this first — it shifts the outline from "what content fits" to "what should the audience remember 10 seconds after each slide." 2. `references/medical_presentation_templates.md` — Section structure, slide counts, and design seeds for the 5 contexts: journal club, grand rounds, conference talk, lecture, and academic lecture multi-paper survey. Pick the matching template after Phase 0 inputs are collected, then customize. 3. `references/slide_visual_styles/` — visual style specs (color palette, typography, layout grid, slide-type templates) callable from any of the 5 context templates. Currently available: nature_lancet.md (Nature/Lancet aesthetic — white background, navy primary, coral accent, Inter/Pretendard). Default for academic lectures per ~/.claude/rules/academic-lecture-style.md. Paired with the generic builder templates/build_pptx_nature_lancet.py and the PDF figure extractor scripts/extract_pdf_figures.py.

These two files mirror the entry-point pattern used in make-figures/references/design_principles.md (Step 1 "Specify"). Both skills share the same Reynolds / Knaflic / Tufte foundations — slide-level (this skill) and figure-level (make-figures) are companions, not duplicates.

Required Inputs

Before starting, collect these from the user:

Paper Analysis

Read the paper and produce a structured analysis:

## Paper Analysis

### Citation
[Full citation with DOI]

### Background
- What gap does this paper address?
- What was known vs. unknown before this study?

### Study Design
- Type: [RCT / cohort / case series / meta-analysis / etc.]
- Subjects: [n, inclusion/exclusion]
- Methods: [key methodological choices]
- Primary outcome: [what was measured]

### Key Results
1. [Finding 1 with effect size and CI/p-value]
2. [Finding 2]
3. [Finding 3]

### Patient/Case Summary Table
[If applicable — structured table of individual cases or subgroups]

### Limitations
1. [Limitation 1]
2. [Limitation 2]

### Significance
- Why does this matter?
- What changes because of this paper?

Slide Outline

Create a slide-by-slide outline with time allocation:

## Slide Outline ([N] slides, [M] minutes)

| # | Title | Time | Key Content |
|---|-------|------|-------------|
| 1 | Title slide | 0:30 | Paper citation, presenter |
| 2 | Context / Prior sessions | 1:00 | How this connects to prior knowledge |
| 3 | Background | 1:30 | The gap this paper fills |
| ... | ... | ... | ... |
| N | Take-home messages | 0:30 | 3-5 key points |

Gate: User approves outline before proceeding.

---

Phase 1: Supporting Research

Search Strategy

Find references that strengthen the presentation:

1. Follow-up studies — Has the main finding been replicated or extended? 2. Clinical trial data — Large-scale data that contextualizes the findings 3. Review articles — Authoritative summaries that frame the topic 4. Contradicting evidence — Important for balanced Q&A preparation

Efficiency rule: Limit supporting references to 5-8 total. Only search categories that the approved outline (Phase 0) actually requires. Skip categories not needed for the presentation type (e.g., skip clinical trials for a methods-focused paper).

Selection Criteria

Do NOT summarize every paper found. Extract only:

• Specific data points needed for slides (incidence rates, OR/HR, AUC values)
• Findings that directly support or challenge the main paper
• Context that helps the audience understand significance

Output

## Verified References

### Main Paper
1. [Citation] — PMID: XXXXX, DOI: XX.XXXX/XXXXX

### Supporting References
2. [Citation] — PMID: XXXXX
   → Used for: [specific data point or context]
3. [Citation] — PMID: XXXXX
   → Used for: [specific data point or context]

### Key Data for Slides
- [Statistic 1]: [value] — Source: [Ref #]
- [Statistic 2]: [value] — Source: [Ref #]

Every reference must have a verified DOI or PMID. Mark unverified references with [UNVERIFIED].

---

Phase 2: Script & Content

Speaker Script

Draft a complete speaker script with these requirements:

1. Language: User's preferred language for narration; English for technical terms 2. Audience adaptation: Adjust explanation depth based on Phase 0 audience profile

• For mixed audiences: add one-line plain-language explanations for specialty-specific terms
• Example: "FLAIR sequence — an MRI technique that suppresses fluid signal to highlight edema"

3. Pronunciation guide: Include native-language pronunciation for drug names, abbreviations

• Example: "lecanemab (leh-KAN-eh-mab)" or local equivalent

4. Timing markers: Note approximate time per slide 5. Transition phrases: Connect each slide to the narrative arc

Structure

## Speaker Script

### Slide 1: Title (0:30)
"[Opening — introduce yourself and the paper]"

### Slide 2: Context (1:00)
"[Connect to prior knowledge or clinical relevance]"

...

### Slide N: Take-home Messages (0:30)
"[Summarize 3-5 key points. Thank audience. Invite questions.]"

Extension Section (Optional)

Only include if user requested in Phase 0. Examples:

• AI/computational research directions stemming from the paper
• Clinical practice implications
• Policy or guideline implications
• Connections to the user's own research

Gate: User reviews script before proceeding.

---

Phase 3: Slides & Notes

Two Modes

Mode A: Generate new slide deck

Generate a fully-editable PPTX from structured inline data using python-pptx. Two canonical template libraries:

• ${CLAUDE_SKILL_DIR}/references/generate_pptx_templates.py — generic T_lead /

T_text / T_table / T_image_right / etc. templates with smoke-tested main(). Use for journal club, grand rounds, conference talk, and short paper talks.

• ${CLAUDE_SKILL_DIR}/templates/build_pptx_nature_lancet.py — Nature/Lancet visual

style (white + navy + coral, Inter/Pretendard, 47-slide academic lecture proven). Use for academic lecture multi-paper survey (template #5). Functions: new_presentation, add_title_slide, add_toc_slide, add_section_divider, add_transition_slide, add_content_slide, add_glossary_slide, add_closing_slide, plus fix_app_xml() helper. Style spec: references/slide_visual_styles/nature_lancet.md.

For lecture decks pulling figures from PDFs (rather than from /make-figures output), use ${CLAUDE_SKILL_DIR}/scripts/extract_pdf_figures.py — pdftoppm + PIL crop with normalized (0–1) box coordinates. Supports both single-crop CLI and YAML batch config.

After raw extraction, run ${CLAUDE_SKILL_DIR}/scripts/trim_caption.py to auto-remove journal headers / figure captions / surrounding whitespace so that only the figure body remains — the Adobe-Acrobat-crop equivalent in automation. The script uses horizontal-projection segmentation plus text-band detection (height + density + gap + line-pattern signature) and preserves multi-panel figures intact:

python3 "${CLAUDE_SKILL_DIR}/scripts/trim_caption.py" \
  --in-dir  figures/extracted \
  --out-dir figures/cropped

Handles four common journal layouts: top running-head bar, bottom multi-line caption (sparse text), bottom caption fused with figure body (no clear gap, detected via narrow dark/light alternation), and multi-row tables with footnotes (footnote cut, table rows preserved). No tesseract / OCR dependency — Pillow + numpy only. Verified on 12-figure academic deck (80–95% height retention; captions, journal banners, and CellPress-style headers all removed). When the deck slot expects only the figure body (default for build_pptx_nature_lancet.py), point FIG_DIR at the cropped output dir.

Word-boundary aware markdown parser (mandatory for HLA-rich decks)

When the build script parses inline **bold** / *italic* markers in slide body or speaker notes, the italic rule must use word-boundary lookahead / lookbehind so asterisk-bearing scientific tokens (HLA alleles like DRB1*07:01, HLA-A*02:01, SNP IDs, footnote markers) are not eaten as italic delimiters:

import re
pattern = re.compile(
    r"(\*\*(?:(?!\*\*).)+?\*\*"                           # bold; inner single * allowed
    r"|(?<![A-Za-z0-9])\*[^*\n]+?\*(?![A-Za-z0-9]))"      # italic (word-boundary)
)

Two regex tricks together: 1. Italic with boundary: (?<![A-Za-z0-9]) and (?![A-Za-z0-9]) reject * adjacent to alphanumerics, so DRB1*07:01 is left intact. 2. *Bold tolerates inner single `**: (?:(?!\\).)+? allows *DRB104:02**` (HLA allele inside bold) to match as a single bold span.

Without these, a naive \*[^*]+\* italic pattern silently corrupts every HLA allele in the deck. Add the regex to add_styled() (or equivalent) in every Nature/Lancet-style build script.

Pronunciation auto-augment for non-native presenters

For decks where the presenter is uncomfortable with English pronunciation of acronyms, author names, drug names, or gene symbols, append a per-slide [ Pronunciation ] section to the speaker notes (audience sees nothing — only Presenter View). Use ${CLAUDE_SKILL_DIR}/scripts/inject_pronunciation_notes.py:

python3 "${CLAUDE_SKILL_DIR}/scripts/inject_pronunciation_notes.py" \
  input.pptx output.pptx \
  --dict pron_dict.yaml \
  --header "[ 발음 ]"            # or any header you like

The script:

• Loads a YAML/JSON PRON_DICT (term → [reading, full_name]) supplied by

the caller. The dict is domain-specific — assemble it for your audience (Korean readings, French readings, Spanish readings, etc.).

• Uses word-boundary regex (?<![A-Za-z0-9_]) … (?![A-Za-z0-9_]) so

short acronyms (e.g. AE, OR) only match when standalone, never inside other words.

• Recognizes allele-style tokens via a separate regex

(\b(?:HLA-)?[A-Z]{1,5}[0-9]?\*[0-9]{2}:[0-9]{2}\b by default) and synthesizes their reading from the base allele entry in the dict.

• Skips slides that already contain the header (idempotent — safe to re-run).

Realistic yield on a 47-slide academic deck: ~38 slides receive a section, ~300 total term entries, 5–10 per annotated slide. Transition and divider slides have empty notes and are auto-skipped.

Speaker notes statistics density

When the slide body already shows exact OR / 95% CI / p-value, the notes should NOT repeat the same numbers — the presenter ends up reading statistics aloud and the audience cannot keep up. Notes should be a narrative (key anchors + one-line "see the slide body for the exact numbers" reminder), not a numeric listing.

Quick measurement to spot dense slides during QC:

import re
text = slide.notes_slide.notes_text_frame.text.split(pron_header)[0]
n_char = len(text)
n_stat = len(re.findall(r"\b(?:OR|p|CI)\s*[=<>]?\s*\d|\d+\.\d+|\d+%|×10", text))
needs_compression = n_char > 1000 and n_stat >= 5

Rule of thumb: 700–1,000 chars + 0–2 stat tokens is fine (30–60-second narrative). >1,000 chars + ≥5 stat tokens → compress to narrative tone and point at the slide body. Exact numbers belong in the slide body and footnotes (SSOT), not the notes.

Sharing-ready notes-stripped variant

After the presentation, when the deck is shared with the audience (e.g. a professor asking for the slides), the speaker notes typically contain presenter-only material — second-language narrative, pronunciation hints, self-referential reminders ("Prof. ○○ will likely ask about …"). Stripping notes is mandatory before circulation. Use ${CLAUDE_SKILL_DIR}/scripts/strip_notes_for_sharing.py:

python3 "${CLAUDE_SKILL_DIR}/scripts/strip_notes_for_sharing.py" \
  presenter_v9.pptx share/<topic>_<initials>.pptx

The script:

• Clears every slide's notes_text_frame (idempotent, slide body and

figures untouched).

• Re-writes docProps/app.xml with the correct Slides= and Notes=

counts so PowerPoint Mac does not show its repair dialog (see also the app.xml canonical fix in pptx-mac-compatibility.md §5).

• Verifies that zero notes characters remain.

Recommended 3-file sharing package (filename pattern <topic>_<initials>):

• <topic>_<initials>.pptx — notes-stripped variant for slide reuse
• <topic>_<initials>.pdf — same deck, PDF for environment-agnostic

preview (LibreOffice --convert-to pdf automatically drops the cleared notes pages)

• <topic>_<initials>_references.zip — optional bundle of the reference

PDFs; if it exceeds the email attachment limit, send a Google Drive link.

In the cover email, mention the PPTX is included specifically so the recipient can reuse individual slides if useful.

Architecture

inline structured data (lists/dicts in build_*_slides())
    ↓ template functions (T_lead / T_text / T_table / ...)
editable PPTX with native text frames (selectable, restyleable in PowerPoint)

Three rules that keep slides stable:

1. No markdown parsing. Every slide is a function call with explicit inline data. 2. No `cur_top` cumulative position tracking. Use the fixed coordinate zones below — cur_top accumulates rounding errors and breaks layout after ~10 slides. 3. No Marp. Marp renders to images; the deck becomes uneditable and reviewers cannot copy text or restyle.

Slide-type templates

Figure source formats (when consuming /make-figures output)

When the deck pulls figures from analysis/figures/ produced by /make-figures:

• Preferred for slides: PNG at ≥300 dpi. python-pptx add_picture() handles this directly. Set img_pct (template T_image_right) so the figure occupies ≥40 % of slide width on a 13.33 × 7.5-in widescreen layout.
• Vector source available: prefer PDF only if the slide will be projected at >1080p or printed as a handout — convert PDF → PNG at the target DPI (pdftoppm -r 300 input.pdf out_prefix) before insertion, because python-pptx PDF embedding is unreliable across PowerPoint versions.
• Forbidden: TIFF (Mac PowerPoint silently drops it — see Mac compatibility checklist below); JPEG for line art (compression artifacts on diagonal lines); raw SVG (PowerPoint Mac handles it inconsistently).
• Caption / legend: re-draft for spoken-narration context, not the journal legend verbatim. The journal legend assumes a reader; the slide caption assumes a listener with 5–10 seconds of attention.

Helpers (used by templates — usually you do not call directly)

Design tokens (defaults — change to fit institution/journal)

NAVY    = #1B2A4A   # title text, section divider background
TEAL    = #0072B2   # subtitle, underline, table header bg, quote bar
ORANGE  = #D55E00   # highlight box border
GRAY    = #333333   # body text
FONT    = 'Apple SD Gothic Neo'   # use a Latin-only font on non-Korean decks

Fixed coordinate zones (16:9 = 13.333" × 7.5")

ML / MR = 0.8"     MT = 0.5"     CW = SW − ML − MR = 11.733"

TITLE_Y = 0.5"    TITLE_H = 0.8"
SUB_Y   = 1.3"    SUB_H   = 0.5"
BODY_Y  ≈ 1.9"    BODY_H  ≈ 5.1"

Build script responsibilities

A from-scratch generation script must:

• Convert TIFF images to PNG before add_picture (Mac PowerPoint silently drops TIFF).
• Apply EXIF transpose to iPhone photos before insertion.
• After inserting/removing slides, sync docProps/app.xml (<Slides>, <Notes>, HeadingPairs, TitlesOfParts) to the actual count, or PowerPoint Mac will raise a recovery dialog on open.
• If you copy <a:srcRect> from another deck, copy the values verbatim — they are 1/1000-percent (cap 100000), never EMU. A unit conversion bug here crops 99% of the image off-slide.
• Print slide count, notes count, file size, and editability check at the end.

Forbidden in Mode A

• ❌ Marp CLI for PPTX (always image-rendered, uneditable).
• ❌ Markdown auto-parsing into slides (layout drifts on every regeneration).
• ❌ cur_top cumulative top tracking (accumulates rounding error).
• ❌ Direct iPhone photo insert without EXIF transpose (rotated 90° in PowerPoint).
• ❌ Using python-pptx from-scratch rebuild to edit an existing deck — see Patch over Rebuild below.

Mac PowerPoint compatibility checklist

PowerPoint Mac is stricter than Windows / Keynote / LibreOffice on OOXML defects. Verify before delivering any deck destined for a Mac viewer:

Validation must run on PDF export AND Mac PowerPoint — neither alone catches all four. PDF misses sp3d outlines and srcRect corruption.

Patch over Rebuild — editing an existing PPTX

When the user supplies an existing deck and asks for surgical edits (textbox width, image crop, font swap, sp3d removal), prefer regex/sed patching of the unzipped XML over regenerating with python-pptx. From-scratch rebuild loses:

• <a:srcRect> image crops
• <a:sp3d> / <a:scene3d> (when intentional)
• Slide master / layout / theme details
• app.xml and core.xml metadata

unzip -q original.pptx -d /tmp/work
python3 -c "
import re; from pathlib import Path
p = Path('/tmp/work/ppt/slides/slide23.xml')
s = p.read_text()
s = s.replace('cx=\"9504720\"', 'cx=\"11200000\"')
p.write_text(s)
"
cd /tmp/work && zip -rq ../patched.pptx . -x '*.DS_Store'

python-pptx is reserved for (a) brand-new decks built via the templates above, or (b) appending speaker notes via slide.notes_slide.notes_text_frame.text. The skill's scripts/inject_speaker_notes.py is the canonical example of (b). It parses inline **bold** / *italic* into run-level styling by default (python-pptx stores text verbatim, so the markers would otherwise show literally in Presenter View — the failure mode pptx-speaker-notes.md warns against); pass --no-markdown for legacy plain text. A reproducible check lives at tests/test_speaker_notes_markdown.py.

Standard structure (10–15 min paper talk)

1. Title slide (T_lead) — paper citation + presenter 2. Background (T_text × 1–2) 3. Study design / Methods (T_text or T_two_col) 4. Key results with figures (T_image_right / T_table × 2–3) 5. Discussion (T_text) 6. Limitations (T_two_col_with_box works well) 7. Take-home (T_text or T_highlight_slide)

Output

Save to output/presentation.pptx. Speaker notes go into the notes pane only — never modify slide design when adding notes.

Step 3.5 — Slide critic (run before delivering deck)

After exporting the PPTX, run the slide critic rubric at references/critic_rubrics/slide.md. Score each slide and the deck-level Mac compatibility checks (Section F) as PASS / PARTIAL / FAIL. Produce concrete edits for every FAIL or PARTIAL item before treating the deck as ready.

Mandatory deck-level checks (cross-link with ~/.claude/rules/pptx-mac-compatibility.md):

# F.22 No TIFF
find ppt/media -iname '*.tif*' || true   # must be empty

# F.23 No 3-D bevel
grep -l '<a:sp3d>' ppt/slides/*.xml      # must be empty

# F.24 app.xml count sync
grep -c '<Slides>\|<Notes>' docProps/app.xml
ls ppt/slides/slide*.xml | wc -l         # must match

# F.25 srcRect bounds (any value > 100000 = bug)
grep -oE '"[0-9]{6,}"' ppt/slides/*.xml | head

Record critic_pass: yes | partial | no and refine_rounds: N in _quick_review.md.

Mode B: Add notes to existing slides (more common)

• Read existing PPTX to understand slide structure and count
• Map speaker script sections to corresponding slides
• Generate inject_notes.py script tailored to the specific presentation

Note Injection Script

Generate a tailored inject_notes.py following the pattern in ${CLAUDE_SKILL_DIR}/scripts/inject_speaker_notes.py. The generated script should contain only the notes dictionary customized for this presentation and the main injection loop from the template.

Critical Rule

Speaker notes are injected without modifying slide design, layout, text, or images. The script only touches the notes pane. Verify by comparing slide content before and after.

---

Phase 4: Q&A Preparation

Question Generation

Generate questions from multiple perspectives:

1. Methodology critics: "Why this design? Why not...?" 2. Domain experts: Deep technical questions about the specific field 3. Generalists: "What does this mean for clinical practice?" 4. Students/trainees: Clarification questions about unfamiliar concepts

Answer Structure

Every answer should follow the pattern:

Acknowledge → Evidence → Conclude

"That's an important limitation. [Acknowledge the concern honestly.]
However, [cite specific supporting evidence — author, year, finding].
So while [restate limitation], [conclude with the paper's contribution despite it]."

Quick Review Sheet

A single-page reference for last-minute review:

## Quick Review

### Must-Know Numbers
| Metric | Value | Source |
|--------|-------|--------|
| [Key stat 1] | [value] | [Ref] |
| [Key stat 2] | [value] | [Ref] |

### Common Pitfalls
- Don't confuse [X] with [Y]
- [Classification A] and [Classification B] are independent frameworks
- Slide says [rounded value], precise value is [exact value]

### Key Takeaways (memorize these)
1. [Point 1]
2. [Point 2]
3. [Point 3]

---

Output File Structure

All outputs go in the user's presentation directory:

{presentation_dir}/
├── _analysis.md              # Phase 0: Paper analysis + outline
├── _references.md            # Phase 1: Verified references + key data
├── _script.md                # Phase 2: Speaker script
├── _qa_prep.md               # Phase 4: Expected Q&A
├── _quick_review.md          # Phase 4: Pre-presentation review sheet + critic_pass record
├── _slide_critic.md          # Phase 3.5: Slide rubric scores per slide
├── inject_notes.py           # Phase 3: Tailored note injection script
├── figures/                  # Extracted paper figures (if needed)
└── reference/                # Supporting paper PDFs (if downloaded)

Cross-skill / Cross-rule integration

This skill composes with adjacent skills and global rules:

---

Constraints

• Never fabricate references. Every citation must be verified against PubMed, DOI, or the PDF itself.
• Never modify slide design when injecting notes. Notes and slides are separate concerns.
• Always ask audience first. Do not start drafting until the target audience is defined.
• Extension sections are opt-in. Do not add AI/clinical/policy sections unless explicitly requested.
• Respect presentation time. Script length must match allocated time (roughly 130-150 words per minute for academic presentations).

Anti-Hallucination

• Never fabricate references. All citations must be verified via /search-lit with confirmed DOI or PMID. Mark unverified references as [UNVERIFIED - NEEDS MANUAL CHECK].
• Never invent clinical definitions, diagnostic criteria, or guideline recommendations. If uncertain, flag with [VERIFY] and ask the user.
• Never fabricate numerical results — compliance percentages, scores, effect sizes, or sample sizes must come from actual data or analysis output.
• If a reporting guideline item, journal policy, or clinical standard is uncertain, state the uncertainty rather than guessing.

Critic Rubric — Slide

Apply this rubric when the generated slide deck (PPTX, PDF export, or rendered preview) is ready for review. Walk every slide and mark each item below as PASS / PARTIAL / FAIL with a one-line justification.

After scoring, produce a list of concrete edits that would resolve every FAIL or PARTIAL item. Return the scored rubric + edit list to the user.

Sources for these rules: slide_design_principles.md (Reynolds, Duarte, Knaflic,

Tufte) and ~/.claude/rules/pptx-mac-compatibility.md. Items marked medical are

drawn from RadioGraphics 2025 ("Winning at the Radiology Podium") and RSNA refresher

course conventions.

---

A. Message clarity (per slide)

1. One key message — the slide conveys exactly one take-home point. A slide with two key messages must be split. 2. Sentence-headline title — the title states the conclusion ("RFA reduced local recurrence by 38%"), not the topic ("Results"). 3. Glance Test ≤ 3 s — a viewer can grasp the slide's meaning within 3 seconds and return attention to the speaker. 4. Reading-time budget matches context — slide content density is appropriate for journal club / grand rounds / conference / lecture (per slide_design_principles.md §2 table).

B. Cognitive load

5. Body text ≤ 6 lines at the slide level (counting wrapped lines, not bullets). 6. Bullet nesting ≤ 2 levels — no 3-indent nested bullets. 7. Charts per slide ≤ 1 — two charts on one slide compete for attention. 8. No decorative animation — animations exist only when they encode a step (build-up of a flow diagram), not for transition flair. 9. No 3-D / drop-shadow / gradient fill unless the third dimension encodes data.

C. Visual hierarchy (Knaflic preattentive)

10. Size hierarchy — title > subtitle > body > caption (≥ 4 pt step between levels). 11. Color carries meaning — accent color used on the one element that matters; rest is neutral. Colors ≤ 3 + 1 accent. 12. Position priority — conclusion / take-home placed at slide top, not bottom. 13. Direct labels on chart series — legend only when ≥ 4 series; otherwise label each curve / bar / wedge directly.

D. Typography

14. Font count ≤ 2 throughout the deck (typically 1 for title, 1 for body). 15. Font size ≥ 18 pt body for slides; ≥ 24 pt for posters; ≥ 28 pt for room sizes

100 seats.

16. Single Korean font when the deck mixes Korean + English (Apple SD Gothic Neo or Pretendard); avoid system Times-fallback for CJK text. 17. Capitalization consistency — sentence case OR title case throughout, not mixed.

E. Native objects (editability)

18. Charts are native PowerPoint objects — double-clicking opens the data table, not an image preview. (Co-authors must be able to update numbers in revision.) 19. Tables are native PowerPoint tables — not pasted screenshots from the paper. 20. Flow diagrams are grouped shapes — not embedded make-figures PNG exports (PNG is acceptable only for visual abstracts and dataset-flow figures intended as one-shot graphics). 21. Images are PNG ≥ 300 dpi or vector PDF — TIFF prohibited (Mac PowerPoint silently drops TIFF — see pptx-mac-compatibility rule).

F. Mac PowerPoint compatibility (deck-level)

Run this once on the final deck before sending to a Mac viewer:

22. No TIFF in `ppt/media/` — find ppt/media -iname '*.tif*' returns empty. 23. No `<a:sp3d>` in slide XML — grep -l '<a:sp3d>' ppt/slides/*.xml returns empty (3-D bevel renders as red outline only on Mac). 24. `app.xml` count synced — <Slides>, <Notes>, HeadingPairs, and TitlesOfParts match the actual file count (mismatch triggers Mac PowerPoint recovery dialog). 25. No `<a:srcRect>` value > 100000 — values are in 1/1000-percent (cap 100000); a unit-conversion bug crops 99% of the image off-slide.

G. Medical presentation specifics

26. Slide pacing — approximately 1 slide per minute of presentation time. A 30-min talk uses 25–30 content slides (excluding title, recap, Q&A buffer). Lectures may pace slower (1 slide / 2 min) when teaching concepts. 27. Background slides ≤ 3 — extended background belongs in the discussion or in backup slides, not in the opening. 28. PICO frame — for evidence-based presentations (journal club, evidence reviews), Population / Intervention / Comparison / Outcome stated explicitly on one slide. 29. Limitations slide present — every paper-driven medical talk has a limitations slide; absence is grounds for FAIL. 30. Strengths / weaknesses 2-column — journal club decks must contrast study strengths and weaknesses (typically a 2-column slide). 31. Statistical vs clinical significance — when a p-value is shown, a follow-up framing of clinical significance (effect size, NNT, absolute risk reduction) is present on the same slide or the next. 32. Disclosure / COI on title slide (or slide 2) — institutional requirement for most medical conferences.

H. References and AI policy

33. Inline citation at point-of-use — [Smith 2024] placed on the slide where the fact appears, not consolidated on a single end-slide. 34. Reference list slide ≤ 1 — if mandated by venue, capped to one slide; otherwise move full bibliography to a handout or supplementary. 35. AI-image policy verified — if any visual abstract / Central Illustration is AI-generated, the target audience/journal AI policy is checked (~/.claude/rules/journal-ai-image-policies.md). JACC family prohibits without EIC permission; Radiology family allows with disclosure.

I. Q&A readiness (Phase 4 cross-check)

36. Backup slides exist for foreseeable methodology questions (≥ 3 anticipated questions covered). 37. Quick Review sheet present — 1-page reference of must-know numbers, common pitfalls, and key takeaways (per SKILL.md Phase 4).

---

Scoring output format

## Critic report (slide deck, round T)

### Slide-by-slide

| Slide # | Title | Item | Score | Note |
|---------|-------|------|-------|------|
| 1 | Title | A.32 disclosure | FAIL | No COI line |
| 5 | Methods | A.2 sentence-headline | PARTIAL | Reads "Methods" — convert to "Cohort: 142 RFA + adjuvant chemo" |
| 12 | Forest plot | C.13 direct labels | FAIL | Legend has 5 series but plot has only 3 |
| ... | | | | |

### Deck-level

| Item | Score | Note |
|------|-------|------|
| F.22 No TIFF | PASS | All images are PNG |
| F.23 No sp3d | FAIL | slide14.xml + slide17.xml have 3-D bevel — strip via regex |
| G.26 Pacing | PASS | 27 content slides for 30-min talk |
| G.29 Limitations | FAIL | Missing — add before take-home |
| ... |

### Required edits before next render
1. Slide 1: Add "Disclosure: No COI" line below presenter name.
2. Slide 5: Rewrite title from "Methods" to one-sentence summary of the cohort.
3. Slide 12: Replace 5-entry legend with direct labels on each curve.
4. Strip `<a:sp3d>` from slide14.xml and slide17.xml.
5. Insert Limitations slide between current slides 24 and 25.

### Overall verdict
[ ] PASS — ready for delivery
[ ] REFINE — items above must be fixed before next round

Record critic_pass: yes | partial | no and refine_rounds: N in _quick_review.md for this presentation after the final round.

Medical Presentation Templates

Section structure, slide counts, and design tone for the four medical presentation contexts handled by this skill. Each template is a starting outline — not a rigid template. Adapt to specific paper, audience, and time slot.

Sources: RadioGraphics 2025 ("Winning at the Radiology Podium," DOI 10.1148/rg.250003);

RSNA Refresher Course Committee guidelines; UAB Physician Scientist Development Office

journal club framework; ASHP journal club template; medsci-skills internal patterns from

2026-04 grand rounds and journal club decks.

>

Triggered from: SKILL.md Phase 0 (Outline) — pick the template matching the

presentation context, then customize. Cross-reference slide_design_principles.md (§2

reading-time budget table) and critic_rubrics/slide.md Section G.

---

1. Journal club (paper-driven, peer trainees)

Audience: residents / fellows + 1–2 faculty. Same-specialty cohort. Time: 20–30 min total (10–15 min present + remainder discussion). Tone: Methodologically rigorous, critical, peer-collegial. Slide count: ~20–25 (1 slide / 30–60 s).

Section structure

Design seed

• Color: Institutional accent (1) + neutral grays (2). No 3-D, no decorative animation.
• Typography: Sans-serif (Helvetica / Calibri / Pretendard for KR). 28 pt title /

18–20 pt body.

• Figures: Use figures from the paper directly (cite slide-corner). Do not re-render

unless adding emphasis annotations.

• Speaker note density: 5–10 s of spoken content per slide.

Pre-presentation checklist

• [ ] Faculty review of paper selection (per UAB framework — verify educational

significance, relevance for practice, validity).

• [ ] Read paper independently before reading the abstract (avoid author-framing bias).
• [ ] Allocate ~50% of total time to discussion, not lecture.
• [ ] Verify submission_safe: true is not needed (journal club ≠ submission).

---

2. Grand rounds (mixed specialty / mixed seniority)

Audience: Multi-specialty, mixed trainee + attending. May include non-radiologists for imaging-themed rounds. Time: 45–60 min total (35–45 min present + remainder Q&A). Tone: Educational, broadly accessible, single-take-home framing. Slide count: ~30–45 (1 slide / 60–90 s, with worked examples).

Section structure

Design seed

• Color: Institutional palette + 1 accent for highlights. Avoid >3 colors per slide.
• Typography: Title 32 pt / body 20–22 pt (room may seat ≥ 100).
• Figures: Mix of paper figures, original imaging, and SMART Servier anatomical icons.

AI-image disclosure if anything is AI-generated.

• Worked examples: Build-up animations acceptable (sequential reveal of imaging

findings) — but only when the build encodes a teaching step.

Pre-presentation checklist

• [ ] Learning objectives written with measurable verbs (avoid "understand"; prefer

"identify," "differentiate," "recommend").

• [ ] Imaging consent / de-identification verified for all cases.
• [ ] Disclosure slide includes industry relationships, advisory roles, prior publications

cited.

• [ ] Mentor pre-review completed (typical 1–2 weeks before).

---

3. Conference talk (RSNA / ECR / KCR / society meeting)

Audience: Specialist peers; high domain expertise. Time: 7–12 min present + 2–3 min Q&A (oral session standard). Tone: Tight, data-dense, focused on contribution. Slide count: ~10–14 for 10-min talk; ~14–18 for 12-min (1 slide / 45–60 s).

Section structure

Design seed

• Color: One accent for primary result; everything else neutral.
• Typography: ≥ 24 pt body throughout (auditorium projection).
• Figures: Re-render journal figures for slide context (drop methodological detail,

enlarge axis labels, direct-label series). RSNA / ECR projection rooms project at 1080p — vector PDFs preferred.

• Speaker pace: 130–150 words/min for academic delivery; rehearse to ensure last

slide arrives at 9:30 of a 10-min slot.

Pre-presentation checklist

• [ ] Abstract was accepted for oral (not poster) — confirm slot type and time.
• [ ] Audience response system not used (oral sessions don't typically include ARS).
• [ ] Submitted disclosure to society (society disclosures often separate from manuscript

COI).

• [ ] Backup slide stack for top-3 anticipated questions ready.

---

4. Lecture / didactic (residents, course material)

Audience: Trainees in structured curriculum (radiology residents, fellows, medical students). Time: 50–60 min (full lecture slot). Tone: Pedagogical, build-up, recap-friendly. Slide count: ~30–40 (1 slide / 90–120 s, with worked examples and recap slides).

Section structure

Design seed

• Color: Course palette consistent across all lectures in series.
• Typography: Title 28 pt / body 20 pt. Build-up animations acceptable for stepwise

concept introduction.

• Figures: Mix of textbook diagrams, original imaging, SMART Servier icons. Reusable

asset library across lecture series (per pptx-mac-compatibility rule — single asset source).

• Recap pattern: At each section end, a 1-slide recap before moving on. Helps

trainees with weaker prior knowledge.

• Self-assessment: Use audience-response polling if classroom supports it; otherwise

show MCQ → pause → reveal answer + explanation slide.

Pre-presentation checklist

• [ ] Lecture position in curriculum confirmed (not duplicating prior lecture content).
• [ ] Self-assessment questions mapped to learning objectives.
• [ ] Pre-reading assignment matches the lecture's prerequisites.
• [ ] Course-level branding consistent (logo, color, font).

---

5. Academic lecture — multi-paper survey (graduate course)

Audience: Graduate students from mixed disciplines (basic science, clinical, imaging, AI). Cross-disciplinary background; a single specialty's jargon will not land for everyone. Time: 30–45 min present + 5–10 min Q&A. Tone: Narrative arc across multiple papers; one lighthouse paper per section anchors the discussion, surrounding papers add nuance. Slide count: ~35–50 (1 slide / 45–75 s, with section dividers + transitions).

Section structure — 4-act narrative

Design seed

• Visual style: Nature/Lancet — see slide_visual_styles/nature_lancet.md
• Body language: English only (per ~/.claude/rules/academic-lecture-style.md §1)
• Speaker notes: narrative, in the user's preferred language (English by default; a Korean narrative register is supported for Korean presenters), 150–300 chars / slide
• Section dividers: full-bleed #121D36 background + coral accent strip + {N} MIN

badge bottom-right

• Transition slides: large coral curly quotes (72pt) flanking a 36pt white question
• Lighthouse paper anchor: each section names one anchor paper in the divider

subtitle; section content slides reference it in eyebrow text

Pre-presentation checklist

• [ ] All -- em-dashes verified rendering correctly on Mac PowerPoint
• [ ] TIFF images converted to PNG (per pptx-mac-compatibility.md §1)
• [ ] <a:sp3d> 3D bevel = 0 (per pptx-mac-compatibility.md)
• [ ] app.xml Slides/Notes counts patched (per pptx-mac-compatibility.md §5)
• [ ] Speaker notes parsed for **bold** / *italic* markdown (per §4)
• [ ] 40-min timing rehearsal completed (longest section flagged)
• [ ] Glossary slide present if audience is multidisciplinary
• [ ] Each section has ≥1 lighthouse paper figure (not text-only)
• [ ] PDF export visually verified (figure positions, font fallback)
• [ ] Mac PowerPoint actual-rendering check (PDF alone misses some defects)

Reference implementation

• Build script: templates/build_pptx_nature_lancet.py (this skill) — add_title_slide,

add_toc_slide, add_glossary_slide, add_section_divider, add_transition_slide, add_content_slide, add_closing_slide + fix_app_xml helper

• Style spec: slide_visual_styles/nature_lancet.md
• Example deck: 47-slide academic lecture (4-section narrative across 17 lighthouse papers

+ supporting references), Mac PowerPoint verified

---

Mapping to SKILL.md phases

Cross-references

• slide_design_principles.md — design foundations (Reynolds, Duarte, Knaflic, Tufte)
• slide_visual_styles/nature_lancet.md — Nature/Lancet color/typography/layout spec
• templates/build_pptx_nature_lancet.py — generic builder for Template #5
• critic_rubrics/slide.md — per-slide quality scoring
• workflow-checklist.md — end-to-end Phase 0 → 4 production checklist
• ~/.claude/rules/academic-lecture-style.md — English body + Korean notes + Nature/Lancet (global default)
• ~/.claude/rules/pptx-mac-compatibility.md — TIFF / sp3d / app.xml / srcRect / Inches EMU / markdown notes
• ~/.claude/rules/multidisciplinary-presentation.md — glossary slide + intuition box pattern
• ~/.claude/rules/journal-ai-image-policies.md — AI-image policy (visual abstract /

Central Illustration cases)

Effective Slide Design for Medical Presentations

Primary sources (cite when invoking specific principles):

- Reynolds G., Presentation Zen: Simple Ideas on Presentation Design and Delivery, 3rd ed.

New Riders, 2019. — Simplicity / Clarity / Restraint / Harmony.

- Duarte N., Slide:ology: The Art and Science of Creating Great Presentations. O'Reilly, 2008.

— Glance Test™ (3-second rule); slides as "glance media."

- Knaflic CN., Storytelling with Data: A Data Visualization Guide for Business Professionals.

Wiley, 2015. — Preattentive attributes (size, color, position).

- Tufte ER., The Cognitive Style of PowerPoint: Pitching Out Corrupts Within, 2nd ed.

Graphics Press, 2006. — Chartjunk; data-ink ratio; bullet-point critique.

>

Medical-presentation specific:

- Bhargava R, et al. "Winning at the Radiology Podium: First-timer's Guide to Crafting and

Delivering a Memorable Conference Presentation." RadioGraphics 2025; DOI 10.1148/rg.250003.

- RSNA Refresher Course Committee guidelines (case-based, audience-response interactive format).

>

Triggered from: SKILL.md Phase 0 (Outline) and Phase 3 (Slides). Read this file before

sketching the slide outline — it shifts focus from "what content fits" to "what should the

audience remember 10 seconds after each slide."

Most slide-design guidance focuses on aesthetic templates (color, font, layout). This file adds a communication-first layer: how the audience parses a slide under time pressure, what they retain, and how the slide supports the speaker rather than replacing them.

---

The 5 design principles (read in order)

1. Identify the key message (most important)

Before opening PowerPoint or python-pptx, write one sentence describing what the slide must convey. If you cannot, the slide is premature. Examples:

• "Adjuvant RFA reduces local recurrence vs. surgery alone (HR 0.62)."
• "FLAIR sequence highlights the lesion that T2 missed."
• "The proposed CAD model maintains AUC across three external sites."

Pin that sentence at the top of the slide outline (Phase 0 of SKILL.md). Every bullet, figure, color, and animation exists to support that one sentence; anything that does not should be removed or moved to a separate slide.

A slide with two key messages is two slides.

2. Audience-aware reading-time budget

Different presentation contexts allow different inspection time per slide. Set the reading-time budget in Phase 0, then design backwards.

The same paper presented at journal club and at RSNA needs two different decks — the RSNA version usually drops half of what the journal-club version contains.

3. The Glance Test (Duarte) — 3-second rule

After each slide is drafted, perform the Glance Test™:

*Can the audience grasp the slide's meaning within 3 seconds, then return their attention

to the speaker?*

If a slide fails the test, the audience is reading the slide instead of listening to the speaker. People cannot read and listen simultaneously; one channel wins, and on slides the reading channel wins. Slides are glance media, not reading documents.

Failure modes that break the Glance Test:

• ≥ 6 lines of body text (audience starts reading paragraph-style).
• Two competing visuals (eye does not know where to land).
• A title that does not state the conclusion (forces reading the body to find the point).

Fix patterns:

• Convert the slide title from a topic ("Methods") to a sentence-summary ("Inclusion required

≥ 1 cm tumor on MRI").

• Split into 2 slides if the message is genuinely two-part.
• Replace bullets with one annotated figure.

4. Reduce cognitive load (Reynolds + Tufte)

Working memory holds ~7 items, and a slide must compete with the speaker's voice. Design under that ceiling.

Tufte's data-ink ratio applies: every pixel should encode either content or essential context. 3-D bars, drop shadows, gradient fills, and rotated axis labels add ink without data — remove them. The Columbia Space Shuttle disaster slide is the canonical counterexample: the O-ring failure data was buried in small type on a crowded slide.

5. Native objects, not images

Charts, tables, and diagrams must be editable native PowerPoint objects, not exported images. This is non-negotiable in academic medicine because:

• Co-authors edit numbers in revision rounds.
• Mac PowerPoint renders TIFF/SVG inconsistently (see pptx-mac-compatibility rule).
• Reviewers / session chairs may project the deck on lower-resolution displays where

rasterized charts pixelate.

Specifically:

• Excel-linked charts → double-click opens the data table.
• Tables → native PowerPoint tables, not screenshots.
• Flow diagrams → grouped shapes, not a make-figures PNG export embedded as image.
• Single fonts (Pretendard / Apple SD Gothic Neo for Korean decks; Helvetica/Calibri for

English-only) — declared once at the deck level, not per-slide.

Allowed exceptions: photographs, anatomical illustrations (SMART Servier, NIAID BioArt), medical imaging frames (CT/MRI slices) — all rasterized by nature.

---

Preattentive attributes (Knaflic) — directing audience attention

Preattentive attributes are visual properties the brain processes automatically (in <250 ms, before conscious attention). Use them to guide the audience to the part of the slide that matters.

Two purposes: (1) draw immediate attention to a key element; (2) build a visual hierarchy that walks the eye through the content.

Anti-pattern: applying preattentive emphasis to everything — when 5 things are bold, red, and boxed, none of them stand out.

---

Slide-anti-patterns (cumulative — drawn from this skill's critic rubric)

These compose with the more granular checks in critic_rubrics/slide.md:

• Topic-titled slide — title says "Results" instead of stating the result. Fix: convert

to a sentence-headline ("RFA reduced local recurrence by 38%").

• Wall-of-text — > 6 body lines, full sentences. Fix: replace with one annotated figure

+ the take-home sentence at top.

• Bullet-cascade — 3-level nested bullets. Fix: split into 2 slides or use a 2-column

layout.

• Chart junk — 3-D bar, gradient fill, rotated axis label, secondary y-axis with no

shared scale. Fix: 2-D, single y-axis, direct labels.

• Image-of-table — pasted screenshot of a table from the paper. Fix: rebuild as native

PowerPoint table; keep the paper-table image for backup slides only.

• Reference-flood — citation list slide listing 15 references at 8 pt. Fix: cite

inline at point-of-use ([Smith 2024]); full bibliography in handout/supplementary.

• Logo-tax — institutional logos on every slide masking content. Fix: title slide and

closing slide only; corner-only on intermediate slides if mandated.

• AI-image without disclosure — AI-generated illustration on a slide destined for a

journal that prohibits AI imagery. Fix: see journal-ai-image-policies rule before building visual abstract / Central Illustration.

---

Cognitive load checklist (Phase 3 quick scan)

Before exporting to PPTX, walk every slide:

• [ ] One-sentence key message stated at slide top (sentence-headline).
• [ ] Body text ≤ 6 lines.
• [ ] Glance test ≤ 3 seconds.
• [ ] Colors ≤ 3 + 1 accent; carry meaning, not decoration.
• [ ] Fonts ≤ 2 throughout deck; ≥ 18 pt body for slides; ≥ 24 pt for posters.
• [ ] Charts/tables are native objects (not images).
• [ ] No 3-D / drop-shadow / gradient unless data-driven.
• [ ] Direct labels on chart series; legends only when ≥ 4 series.
• [ ] Same slide works in grayscale (run a convert -colorspace Gray test).
• [ ] AI-image policy verified for target audience/journal context.

If two or more boxes are unchecked, return to Phase 0 outline before continuing.

---

Cross-references

• critic_rubrics/slide.md — quantitative critic checks per slide
• medical_presentation_templates.md — section structure for journal club, grand rounds,

conference, lecture

• workflow-checklist.md — end-to-end Phase 0 → 4 production checklist
• make-figures/references/design_principles.md — figure-level design (this skill is the

slide-level companion; both share Reynolds/Knaflic/Tufte foundations)

• ~/.claude/rules/pptx-mac-compatibility.md — TIFF, sp3d, app.xml, srcRect defects
• ~/.claude/rules/journal-ai-image-policies.md — AI-image policy (JACC prohibits, Radiology

allows with disclosure)

• ~/.claude/rules/manuscript-style-classical.md — heading style for slides paired with

manuscript submission

Slide Visual Style — Nature / Lancet

Design spec for academic lecture, journal club, and conference-talk slides modeled on the Nature / Lancet journal aesthetic. White background, navy primary, restrained coral accent, hairline dividers, two-font sans-serif system.

Triggered from: SKILL.md Phase 3 (Slides & Notes). Pair with

references/medical_presentation_templates.md Template #5 "Academic lecture

(multi-paper)". Reference build script: templates/build_pptx_nature_lancet.py.

---

1. Color palette

Rule of restraint: a content slide uses white + navy + 1 coral accent. Three colors maximum. Avoid full-bleed bright bars, gradients, drop shadows beyond the single 6pt offset on figure tiles.

2. Typography

Font install (macOS):

brew install --cask font-pretendard font-inter

EastAsia attribute must be set on every run that may contain Korean text — Mac PowerPoint falls back to Times New Roman otherwise. See ~/.claude/rules/pptx-mac-compatibility.md §6.

3. Layout grid (16:9, 13.333" × 7.5")

Margins: 0.7" left/right, 0.32" top eyebrow, 0.35" bottom footer. Title-body separation enforced by the 0.6" coral hairline (never a 0.05"-tall bar).

4. Slide type templates

4a. Title slide

• Left navy bar (40k EMU wide × 3.5" tall) at x=0.7"
• "REVIEW LECTURE" eyebrow (14pt, coral, letter-spaced 300)
• 48pt navy bold title
• 18pt italic subtitle (TEXT_SUB)
• Bottom block: 16pt bold navy line ("Course · Professor · Date"), 13pt sub line

("Presenter Name · Affiliation"), separated by 2" navy hairline above

4b. Section divider

• Full-bleed deep navy background (DIVIDER_BG)
• 1.8"-tall coral accent strip at x=1.2", y=3.0", 15k EMU wide
• "SECTION {N}" 18pt coral eyebrow letter-spaced 400
• 52pt white bold section title
• 20pt italic light-navy (#C0CBDC) subtitle
• Bottom-right "{N} MIN" badge (13pt muted)

4c. Transition slide

• Full-bleed deep navy background
• Large coral curly quotes (72pt) flanking a 36pt white bold single-sentence question
• Center-anchored vertically

4d. Content slide

• Eyebrow (small caps topic label) → title + subtitle → coral hairline → bullets

(left if figure, full-width if no figure) → figure tile (shadow offset 0.06", hairline border) → optional figure caption → footnote (right) + page brand (left)

• Bullet markers:
• Main: ▪ (14pt coral bold) + 20pt body text
• Sub (lines prefixed with 2 spaces): — (15pt muted) + 16pt sub text
• Inline **bold** and *italic* markdown is parsed into per-run styling (see

pptx-mac-compatibility.md §4 for the parser pattern)

4e. TOC / outline slide

• Eyebrow "OUTLINE"
• Title "Outline" + sentence subtitle
• Coral hairline
• N rows, each: 22pt coral section number (or · for wrap-up), 22pt navy bold section

title, 13pt sub-text English summary, 12pt muted time badge (right-aligned). Hairline divider between rows.

4f. Glossary slide (optional, for multidisciplinary audiences)

• Tier 1 (top, 4–7 items): disease/concept abbreviations with one-line context
• Tier 2 (bottom, 2-column, 8–12): method/statistics abbreviations with short defs
• See ~/.claude/rules/multidisciplinary-presentation.md §1

4g. Closing slide

• Title "Take-home messages" or "Conclusions"
• 3 bullet maximum
• Optional contact / acknowledgments block at bottom

5. Figure handling

• Aspect-ratio preserving fit inside the figure tile (5.0" × 4.0" max)
• Shadow rectangle (hairline gray) offset +0.06" / +0.06" before image
• Image border: hairline gray (~6000 EMU)
• Caption format: Figure · {caption} (coral "Figure" eyebrow, italic muted caption)

EMU pitfall: compute width/height in inches before wrapping in Inches(). Never Inches(fig_w / aspect_ratio) when fig_w is already a pixel value — see pptx-mac-compatibility.md §7.

6. Speaker notes density

• Narrative speaker notes in the user's preferred language (a Korean narrative register is supported for Korean presenters, per ~/.claude/rules/academic-lecture-style.md §1)
• About 150–300 characters per slide (30–60 seconds spoken)
• Run-level markdown parser for **bold** / *italic* (see

pptx-mac-compatibility.md §4)

• 13pt Pretendard

7. Mac compatibility checklist (run before delivery)

Always validate on Mac PowerPoint + PDF export. PDF alone misses some defects (see pptx-mac-compatibility.md §1–§3).

8. Reference implementation

Canonical build script: templates/build_pptx_nature_lancet.py (in this skill).

Example output: 47-slide academic lecture deck (4-section narrative across multiple lighthouse papers + supporting references), Mac PowerPoint verified.

Cross-references

• ~/.claude/rules/academic-lecture-style.md — global style rule (English body + Korean notes + this design)
• ~/.claude/rules/pptx-mac-compatibility.md — Mac compatibility (TIFF / sp3d / app.xml / Inches EMU / markdown notes)
• ~/.claude/rules/multidisciplinary-presentation.md — glossary slide + intuition box pattern
• references/medical_presentation_templates.md Template #5 — Academic lecture (multi-paper)
• references/slide_design_principles.md — Reynolds / Duarte / Knaflic / Tufte foundations

Presentation Preparation Checklist

Use this checklist to verify completeness at each phase.

---

Phase 0: Init & Outline

Paper Analysis

• [ ] Full citation with DOI/PMID recorded
• [ ] Background gap clearly identified
• [ ] Study design classified (RCT, cohort, case series, etc.)
• [ ] Sample size and key inclusion/exclusion criteria noted
• [ ] Primary and secondary outcomes listed
• [ ] Key results with effect sizes and confidence intervals
• [ ] Limitations enumerated (minimum 3)
• [ ] Clinical/scientific significance stated

Audience Assessment

• [ ] Target audience specialties identified
• [ ] Knowledge level estimated (student / trainee / specialist / mixed)
• [ ] Specialty-specific terms that need plain-language explanation listed
• [ ] Prior session topics noted (for coursework presentations)

Outline

• [ ] Slide count matches presentation time (roughly 1 slide per minute)
• [ ] Time allocated per slide
• [ ] Narrative arc: context → gap → methods → results → implications
• [ ] Opening connects to audience's existing knowledge
• [ ] Closing has clear take-home messages (3-5 maximum)

---

Phase 1: Supporting Research

Search Completeness

• [ ] Follow-up studies searched (has finding been replicated?)
• [ ] Contradicting evidence searched (balance for Q&A)
• [ ] Review articles checked (authoritative framing)
• [ ] Clinical trial data checked (real-world context)

Citation Verification

• [ ] Every reference has DOI or PMID
• [ ] No fabricated citations
• [ ] Unverifiable references marked [UNVERIFIED]

Data Extraction

• [ ] Only presentation-relevant data extracted (not full paper summaries)
• [ ] Data source (reference number) linked to each statistic
• [ ] Numbers cross-checked against original papers

---

Phase 2: Script & Content

Audience Adaptation

• [ ] Technical terms explained at appropriate depth for audience
• [ ] Mixed-audience terms have one-line plain-language additions
• [ ] No unexplained abbreviations on first use
• [ ] Pronunciation guides for drug names and key abbreviations

Script Quality

• [ ] Script matches allocated time (130-150 words/min for academic talks)
• [ ] Transitions between slides are explicit
• [ ] Opening hooks audience attention
• [ ] Closing reinforces key messages
• [ ] Limitations acknowledged honestly (not hidden)

Optional Extension

• [ ] Extension section included only if user requested
• [ ] Extension connects naturally to main content
• [ ] Extension has concrete evidence (not speculation)

---

Phase 3: Slides & Notes

Note Injection

• [ ] Notes mapped to correct slide numbers
• [ ] Slide content NOT modified (only notes pane)
• [ ] Script verified by running inject script on a copy first
• [ ] Output file opened and spot-checked in presentation software

Content Consistency

• [ ] Numbers in notes match numbers on slides
• [ ] If slides round numbers (e.g., 12% vs 10.9%), notes use precise values
• [ ] Slide titles match script section references

---

Phase 4: Q&A Preparation

Question Coverage

• [ ] Methodology questions (study design choices, limitations)
• [ ] Domain expert questions (deep technical details)
• [ ] Generalist questions (clinical relevance, so-what)
• [ ] Student questions (concept clarification)

Answer Quality

• [ ] Each answer follows: Acknowledge → Evidence → Conclude
• [ ] Limitations acknowledged, not deflected
• [ ] Supporting evidence cited with author/year
• [ ] Answers are concise (30-60 seconds speaking time each)

Quick Review Sheet

• [ ] Key numbers in a single table
• [ ] Common confusion points listed
• [ ] Classification frameworks distinguished (if multiple exist)
• [ ] Precise values noted where slides use rounded numbers