Academic Lectures Skill

Builds structured medical lecture .pptx presentations with speaker notes, evidence-based content, and optional AI-generated visuals. Always run in full mode — no shortcuts.

Step 1: Intake

Use AskUserQuestion to collect all of the following at once:

1. Topic — What is the lecture topic?
2. Audience — Who is the audience? Options: medical students / residents / fellows / attendings / mixed clinical (nurses, techs, other staff)
3. Duration — How long is the talk? (e.g., 20 min, 30 min, 45 min, 60 min) → slide count will be recommended
4. Deadline — When is this due? (date or relative, e.g., "next Friday", "in 2 weeks")
5. Output path — Where should the project folder and final .pptx be saved?
6. Speaker notes style — Bullet-point talking points (default) or full narrative script?
7. References — Will you provide your own? Or should I run a light PubMed search?
8. Images — Should I generate AI illustrations for relevant slides, or will you provide your own?
9. Take-home points placement — Best practice: show take-home points right after learning objectives AND repeat them at the end. Keep this default? (yes/no)

Step 2: Slide Count Recommendation

Recommend based on duration at ~1.5 min/slide:

Duration	Recommended slides
20 min	13–14
30 min	18–20
45 min	28–30
60 min	38–40

Present the recommendation and ask the user to confirm or adjust before proceeding.

Step 3: Project Folder Setup

Create a project subfolder at the user's specified output path:

lecture-[topic-slug]/
├── lecture-brief.md        ← human-editable direction file (user can update anytime)
├── outline.md              ← structured slide-by-slide content
├── progress.md             ← LQS log (phased prep only)
├── resources/
│   ├── references.md       ← collected or user-provided references
│   ├── daily-YYYY-MM-DD.md ← Phase 1 daily collection logs (phased prep only)
│   └── synthesis.md        ← end-of-Phase-1 synthesis (phased prep only)
└── [topic-slug]-lecture.pptx

Slugify the topic: lowercase, hyphens, no special characters (e.g., "Atrial Fibrillation" →

atrial-fibrillation

).

Always create

lecture-brief.md

at setup using the template in

references/scheduled-prep.md

— even for immediate development. It anchors the lecture goals and the user can edit it to redirect at any time.

Step 4: References

User-provided: Ask them to paste references now or drop them into

resources/references.md

. Use Vancouver/AMA numeric citation format.

Light PubMed search: Use WebSearch to find 5–8 key papers. Search query format:

site:pubmed.ncbi.nlm.nih.gov [topic] [relevant subtopics]

. Save results to

resources/references.md

with title, authors, journal, year, PMID. Flag to user that citations must be verified before delivery.

Step 5: Deadline Check → Phased Preparation

Ask the user:

"Do you want to sleep on it — let me collect resources and build the outline over multiple days so the slides are ~80% ready when you sit down to review? Or should I develop the slides now?"

• Sleep on it → activate phased preparation. See

  references/scheduled-prep.md

  for full setup. Uses the deadline provided in intake to calculate phase lengths.
• Develop now → skip to Step 6 immediately.

If the user chose "sleep on it" but gave no deadline or a deadline ≤2 days away, flag that phased prep requires at least 3 days and offer to proceed with immediate development instead.

Step 6: Generate Outline

Build

outline.md

using the medicine-specific slide structure. See

references/slide-structure.md

for full structure, slide types, visual recommendations, and audience calibration guidance.

Key rules:

• Calibrate vocabulary and depth to the specified audience
• Take-home points appear twice: after learning objectives AND at the end (unless user opted out)
• Every slide entry includes: title, bullet content, speaker notes, and image recommendation

Format each slide in

outline.md

:

## Slide N: [Title]
**Type:** [slide type from slide-structure.md]
**Content:**
- bullet 1
- bullet 2

**Speaker Notes:**
- talking point 1
- talking point 2

**Image:** [None | anatomical diagram: description | pathophysiology flowchart: description | data visualization: description | clinical scenario illustration: description]

Step 7: Image Generation

If user opted for AI-generated images:

1. Review

   outline.md

   — identify all slides with a non-None image recommendation
2. Eligible types: anatomical diagram, pathophysiology flowchart, data visualization, clinical scenario illustration
3. Skip: title, agenda, learning objectives, take-home points, references, and pure text/table slides
4. For each eligible slide, craft a detailed Gemini image prompt: medically accurate, clean educational style, appropriately labeled, white/light background
5. Invoke

   baoyu-danger-gemini-web

   skill for each image
6. Save images to

   resources/

   and update the image field in

   outline.md

   with the filename

Step 8: Assemble .pptx

Invoke

anthropic-skills:pptx

to build the final presentation from

outline.md

.

Instructions to pass:

• Ask user for template preference or let pptx skill choose a clean professional academic theme
• Map each slide entry in

  outline.md

  directly to a slide
• Speaker notes go in the notes pane of each slide
• Embed generated images in their respective slides at a visually prominent position
• Save as

  [topic-slug]-lecture.pptx

  inside the project folder

Step 9: Confirm and Summarize

After generation, report:

• Full path of saved .pptx
• Total slide count
• Which slides have AI-generated images
• Any placeholders requiring user attention (missing data, unverified citations, open questions)
• If phased prep was configured: the schedule and what to expect each day

---

Reference Files

• references/slide-structure.md

  — full medicine-specific slide architecture, types, ordering, and audience calibration

• references/scheduled-prep.md

  — phased preparation workflow for deadline-driven lectures