Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate pedagogically-aligned slide decks from educational content using NotebookLM. Use when creating chapter slide presentations with proficiency-calibrated prompts. NOT for static slides or non-educational presentations.
.claude/skills/aiskillstore-notebooklm-slides/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 2841% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -10% | 0% |
bash# 1. Start browser (via browser-use skill) bash .claude/skills/browser-use/scripts/start-server.sh # 2. Navigate to NotebookLM # browser_navigate to notebooklm.google.com # 3. Create notebook, upload sources, generate slides # Use proficiency-calibrated prompts below
| Step | Action | Tool | |------|--------|------| | 1 | Navigate to notebooklm.google.com | browser_navigate | | 2 | Create notebook: "Chapter X: Title" | browser_click | | 3 | Upload ALL sources (lessons + README + quiz) | browser_click | | 4 | Click "Slide Deck" in Studio panel | browser_click | | 5 | Select "Presenter Slides" format | browser_click | | 6 | Paste proficiency-calibrated prompt | browser_type | | 7 | Click "Generate" (wait 5-30 min) | browser_click | | 8 | Review with success criteria | Visual inspection | | 9 | Download PDF | browser_click | | 10 | Move to static/slides/chapter-{NN}-slides.pdf | Bash |
Create inspiring slide deck for absolute beginners (A2 proficiency).
AUDIENCE: Complete beginners with no programming experience.
FRAMEWORK TO EMPHASIZE:
• [Principle 1]: Simple, concrete explanation
• [Principle 2]: Accessible mental model
• [Principle 3]: Encouraging principle
THEMES (with specific data):
1. [Theme with concrete numbers/facts]
2. [Theme with specific example]
3. [Theme with real-world data]
TONE:
• Encouraging (not intimidating)
• Future-focused and opportunity-driven
• Simple language, no jargon
• Action-oriented
<slide_format_requirements>
Generate 12-15 slides. Each slide: 3-5 bullet points as sentences,
NOT paragraphs. Clear headings. Cover all themes.
</slide_format_requirements>
NARRATIVE: problem → transformation → opportunity → action
END WITH: Specific next steps (not "Keep learning!")Create comprehensive slide deck for intermediate learners (B1 proficiency).
AUDIENCE: Learners with [prerequisites]. Ready for [next-level challenge].
FRAMEWORK TO EMPHASIZE:
• [Intermediate concept with practical context]
• [Problem-solving approach]
• [Real-world application pattern]
THEMES (with specific data):
1-5. [Themes with concrete examples]
TONE:
• Professional yet accessible
• Balance theory with practice
• Technical terms with context
• Critical thinking encouraged
<slide_format_requirements>
Generate 15-20 slides. Each slide: 4-6 bullet points.
Include practical examples and case studies.
</slide_format_requirements>
END WITH: Implementation strategies (step-by-step)Create detailed slide deck for advanced practitioners (C1 proficiency).
AUDIENCE: Experienced with [advanced prerequisites].
FRAMEWORK TO EMPHASIZE:
• [Theoretical frameworks and trade-offs]
• [Industry patterns and anti-patterns]
• [Critical analysis and decision-making]
THEMES: [5-7 themes with industry data]
TONE:
• Professional and rigorous
• Nuance and complexity
• Industry-standard terminology
• Analytical and evaluative
<slide_format_requirements>
Generate 20-25 slides. Each slide: 5-7 bullet points.
Include architecture diagrams, decision matrices.
</slide_format_requirements>
END WITH: Production deployment strategies| Gate | Check | Pass | Fail | |------|-------|------|------| | 1. Title | Reflects framework? | "AI Coding Revolution" | "Introduction to AI" | | 2. Language | Matches proficiency? | A2: simple, no jargon | A2 with technical terms | | 3. Themes | All 5-7 covered? | Each theme with data | Themes missing | | 4. Tone | Matches spec? | Encouraging (not academic) | Wrong emotional framing | | 5. Count | Within range? | A2: 12-15, B1: 15-20 | Outside range | | 6. Arc | Progression clear? | problem → action | Random sequence | | 7. Ending | Actionable? | Specific tasks | "Keep learning!" |
Score: 7/7 → Deploy | <7/7 → Iterate with refined prompt
Format: chapter-{NN}-slides.pdf (zero-padded)
bash# Example mv ~/Downloads/"The-AI-Revolution.pdf" \ "apps/learn-app/static/slides/chapter-01-slides.pdf"
Add to chapter README frontmatter:
yaml--- title: "Chapter 1: Title" slides: source: "slides/chapter-01-slides.pdf" title: "Chapter 1: Title" height: 700 ---
Build-time plugin auto-injects PDFViewer before "What You'll Learn".
For 3+ chapters:
Daily limit: 3-5 chapters/day (NotebookLM enforced)
| Issue | Solution | |-------|----------| | Generation stuck >30 min | Check browser console, verify no daily limit message | | Text-heavy slides | Add explicit "3-5 bullets, NOT paragraphs" | | Generic title | Include example engaging title in prompt | | Missing themes | List all themes numbered with specific data | | Daily limit hit | Wait 24h (midnight PT reset), notebooks persist |
| Don't | Why | Do Instead | |-------|-----|------------| | Vague audience | NotebookLM can't calibrate | "A2 beginners with no programming" | | Skip framework | Generic output | Explicit 3-5 principles | | Single-word tone | Ambiguous | "Encouraging (not intimidating)" | | Leave format default | Text-heavy slides | Explicit bullet count | | Vague endings | No student action | Specific next steps |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 10,117 | 2,753 | -73% | 1 | 1 | 0% | 1,513 | 1,993 | +32% | 0 | 0 | — |
case-01 | fail→fail | 5,062 | 5,031 | -1% | 1 | 1 | 0% | 343 | 1,955 | +470% | 0 | 0 | — |
case-02 | fail→pass | 4,961 | 24,434 | +393% | 1 | 1 | 0% | 205 | 6,029 | +2841% | 0 | 0 | — |
case-03 | fail→fail | 18,932 | 3,905 | -79% | 1 | 1 | 0% | 3,605 | 1,908 | -47% | 0 | 0 | — |
case-04 | pass→fail | 6,172 | 4,871 | -21% | 1 | 1 | 0% | 1,064 | 1,926 | +81% | 0 | 0 | — |
case-05 | fail→fail | 2,961 | 4,711 | +59% | 1 | 1 | 0% | 571 | 1,832 | +221% | 0 | 0 | — |
case-06 | pass→fail | 5,287 | 7,506 | +42% | 1 | 1 | 0% | 912 | 1,973 | +116% | 0 | 0 | — |
case-07 | fail→pass | 11,295 | 8,141 | -28% | 1 | 1 | 0% | 1,665 | 2,821 | +69% | 0 | 0 | — |
case-13 | fail→pass | 12,700 | 2,649 | -79% | 1 | 1 | 0% | 1,975 | 2,017 | +2% | 0 | 0 | — |
case-08 | fail→pass | 14,652 | 3,023 | -79% | 1 | 1 | 0% | 2,318 | 2,093 | -10% | 0 | 0 | — |
case-09 | pass→pass | 10,543 | 4,880 | -54% | 1 | 1 | 0% | 1,523 | 2,338 | +54% | 0 | 0 | — |
case-10 | fail→pass | 8,993 | 2,248 | -75% | 1 | 1 | 0% | 1,268 | 1,921 | +51% | 0 | 0 | — |
case-11 | fail→pass | 7,269 | 2,375 | -67% | 1 | 1 | 0% | 1,070 | 1,920 | +79% | 0 | 0 | — |
case-14 | fail→pass | 7,128 | 2,931 | -59% | 1 | 1 | 0% | 1,271 | 2,062 | +62% | 0 | 0 | — |
case-15 | fail→pass | 9,184 | 2,128 | -77% | 1 | 1 | 0% | 1,597 | 1,936 | +21% | 0 | 0 | — |
case-16 | fail→pass | 12,166 | 2,756 | -77% | 1 | 1 | 0% | 1,943 | 2,063 | +6% | 0 | 0 | — |
case-17 | fail→pass | 6,036 | 2,195 | -64% | 1 | 1 | 0% | 932 | 1,931 | +107% | 0 | 0 | — |
case-18 | fail→pass | 7,644 | 1,650 | -78% | 1 | 1 | 0% | 1,027 | 1,858 | +81% | 0 | 0 | — |
case-19 | pass→pass | 11,250 | 5,475 | -51% | 1 | 1 | 0% | 1,815 | 2,473 | +36% | 0 | 0 | — |
case-20 | pass→pass | 10,692 | 5,455 | -49% | 1 | 1 | 0% | 1,676 | 2,458 | +47% | 0 | 0 | — |
case-21 | fail→pass | 10,279 | 1,443 | -86% | 1 | 1 | 0% | 1,658 | 1,785 | +8% | 0 | 0 | — |
case-22 | fail→pass | 10,571 | 2,658 | -75% | 1 | 1 | 0% | 1,623 | 1,940 | +20% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 16 counted toward the lift figure. The other 6 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 16 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.