/presentation-builder
Build and polish presentations using Michal's workshop method + Oren Efraim's rules. Guides users through premise, framing, drafting, polishing, opening/closing, and practice. Works with voice (VoiceLayer) or text.
$ golems-cli skills install presentation-builderUpdated 2 weeks ago
Build great presentations step by step. Based on Michal's Speakers Workshop (TechGym/Melio), Oren Efraim's "Art of Building and Delivering Great Presentations", Uri Alon (Weizmann), and 6+ expert sources.
Full rules reference (local, gitignored):
docs.local/speakers-workshop/presentation-rules.md
Usage
/presentation-builder [workflow]Workflows:
/presentation-builder premise— Find and refine the single premise (Session 1)/presentation-builder build— Frame, draft with AI, polish, slides, opening/closing (Session 2)/presentation-builder review— Review and practice a finished presentation (Session 3)/presentation-builder full— Walk through the entire process start to finish/presentation-builder rules— Show all extracted presentation rules (quick reference)
The Method (3 Sessions)
Session 1: Find the Main Subject (Premise)
Goal: One sentence that captures the entire talk. Everything else flows from this.
Process:
- Ask the user: "What's the talk about? Who's the audience? How long?"
- Brainstorm 3-5 possible angles together
- Narrow to ONE premise — it should be tweetable
- Test: "If the audience remembers only one sentence, what is it?"
Uri Alon's Rule: Every good talk has ONE single premise. Write it FIRST. If a slide doesn't support it, cut it.
Output: A single premise sentence + audience definition + time constraint.
Session 2: Build the Presentation
This is the main work session. Follow this order strictly:
Step 1: Frame the Structure
Use Oren Efraim's 10-minute structure (scales to any length):
- Hook — grab attention (don't start with "about me")
- Problem — what are we solving?
- Solution — the answer
- Why this / Why us — what makes it special?
- Summary — callback to opening
Ask the user to map their content to these 5 sections.
Step 2: Build the First Draft with AI
Help the user write a prompt for AI to generate a first draft:
- Include: premise, audience, time limit, structure, key points per section
- Include: tone (casual/formal), language, any must-include stories or data
- The AI draft is a STARTING POINT, not the final product
Step 3: Polish and Synthesize (Michal's Rule)
Go through the draft and:
- Cut ruthlessly — if it doesn't serve the premise, remove it
- Synthesize — combine overlapping points, tighten language
- One message per slide (Oren Efraim's golden rule)
- Kill bullet points — max 3 per slide, or use images instead
- Big text — what looks good on laptop doesn't look good projected
Step 4: Content/Media (Do This BEFORE Opening/Closing)
For each slide, check:
- Does it convey exactly ONE idea?
- Is the text big enough to read from the back?
- Are bullet points minimized? (max 3, highlight one at a time)
- Colors: max 5, consistent palette
- Code: monospace, large, syntax highlighted, show LESS not more
- Charts: simplified, noise removed, key data highlighted
- Architecture diagrams: progressive reveal, not everything at once
- Numbers: big and prominent (they're proof, not the story)
Oren Efraim: 10-30 seconds per slide. If you need more time, split the slide.
Step 5: Opening and Closing (LAST — Michal's Rule)
Opening is the LAST thing you write. You need to know all your slides before you know the best hook.
5 Hook Types (Oren Efraim):
| Type | How | Example |
|---|---|---|
| Question | Trigger curiosity | "What happens when you build something for yourself and it becomes a system?" |
| Shared Pain | Empathy hook | "Remember deleting the production database?" |
| Personal Story | Create proximity | "Two years ago I couldn't speak in front of anyone..." |
| Stunning Fact | Challenge assumptions | "We see 1,200 presentations a year. We remember 3%." |
| Paint a Future | Optimism hook | "Imagine a world where every talk you give is remembered." |
Closing: The LAST point MUST callback to the opening. Full circle = closure.
Session 2 output: A complete presentation wireframe — structure, polished slides, opening, closing. Ready to practice.
Session 3: Practice and Polish
Goal: Make the presentation as good as possible. Practice, get feedback, refine.
This is where you take the wireframe from Session 2 and turn it into a performance.
Practice Runs
- Practice out loud — don't just skim in your head (Oren Efraim)
- Memorize only the first 30 sec - 2 min — the opening must be smooth and powerful
- After opening, use slides as anchors — each slide = one message, talk naturally
- Record yourself (video) — watching yourself is the fastest improvement
- Time it: 1 minute = ~150 words
- Have water nearby — drinking = strategic pause
If voice is available (VoiceLayer), use voice_ask for practice runs:
- Speak the opening, get feedback
- Time the full run
- Note where energy drops
Review Checklist
- Premise is clear in the first 2 minutes
- Hook grabs attention (not "Hi, I'm X and today I'll talk about...")
- Each slide = one message
- Energy varies: warm (story) → cool (technical) → warm (insight) → cool (data) → warm (close)
- Closing callbacks to opening
- Timed and within limit
- First 30 sec - 2 min memorized cold
Feedback to Give/Look For
- Where did energy drop?
- Which slide felt too long?
- Was the hook compelling?
- Did the closing feel connected to the opening?
- Audience engagement: would you have stayed?
Present to the Group
Show your opening, get live feedback, iterate. This is the final polish before the real event.
Delivery Rules (Quick Reference)
Voice & Pace
- Speak clearly, vary pace — slow for important, faster for setup
- Use methodical pauses — silence after key points is powerful
- Water = strategic pause when rushing or confused
Body
- Stand, don't sit. Use the stage. Hands visible.
- Face the audience, not the screen
- Eye contact with individuals (3-5 sec each), including back rows
Nerves
- Reframe: excitement, not fear (same physical symptoms)
- Breathing: 5 sec in, 5 sec out (Huberman protocol)
- First 30 seconds are the hardest — it gets easier
- Know your first sentences cold
DON'T
- Don't memorize word-for-word — sounds robotic
- Don't apologize ("This is my first time")
- Don't read slides with your back to the audience
- Don't switch languages mid-sentence (unless technical term)
Environment
- Arrive early, check projector/adapter
- Close all notifications (DND mode)
- Know your mic type (Madonna = belt pack)
- Conferences are always cold — wear sleeves
Audience Engagement
- Raise hands / small jokes work in most cases
- Have a comeback if audience doesn't participate
- The Airplane! trick: Plant a mystery number/fact at the start — someone will ask about it at the end (solves "no questions" problem)
Q&A (Uri Alon)
- Listen to the full question
- Repeat it for the audience
- "I don't know" is valid — follow with "but here's what I think..."
- Long/confused questions: "Let me make sure I understand — are you asking...?"
Full SKILL.md source — includes LLM directives, anti-patterns, and technical instructions stripped from the Overview tab.
Build great presentations step by step. Based on Michal's Speakers Workshop (TechGym/Melio), Oren Efraim's "Art of Building and Delivering Great Presentations", Uri Alon (Weizmann), and 6+ expert sources.
Full rules reference (local, gitignored):
docs.local/speakers-workshop/presentation-rules.md
Usage
/presentation-builder [workflow]Workflows:
/presentation-builder premise— Find and refine the single premise (Session 1)/presentation-builder build— Frame, draft with AI, polish, slides, opening/closing (Session 2)/presentation-builder review— Review and practice a finished presentation (Session 3)/presentation-builder full— Walk through the entire process start to finish/presentation-builder rules— Show all extracted presentation rules (quick reference)
The Method (3 Sessions)
Session 1: Find the Main Subject (Premise)
Goal: One sentence that captures the entire talk. Everything else flows from this.
Process:
- Ask the user: "What's the talk about? Who's the audience? How long?"
- Brainstorm 3-5 possible angles together
- Narrow to ONE premise — it should be tweetable
- Test: "If the audience remembers only one sentence, what is it?"
Uri Alon's Rule: Every good talk has ONE single premise. Write it FIRST. If a slide doesn't support it, cut it.
Output: A single premise sentence + audience definition + time constraint.
Session 2: Build the Presentation
This is the main work session. Follow this order strictly:
Step 1: Frame the Structure
Use Oren Efraim's 10-minute structure (scales to any length):
- Hook — grab attention (don't start with "about me")
- Problem — what are we solving?
- Solution — the answer
- Why this / Why us — what makes it special?
- Summary — callback to opening
Ask the user to map their content to these 5 sections.
Step 2: Build the First Draft with AI
Help the user write a prompt for AI to generate a first draft:
- Include: premise, audience, time limit, structure, key points per section
- Include: tone (casual/formal), language, any must-include stories or data
- The AI draft is a STARTING POINT, not the final product
Step 3: Polish and Synthesize (Michal's Rule)
Go through the draft and:
- Cut ruthlessly — if it doesn't serve the premise, remove it
- Synthesize — combine overlapping points, tighten language
- One message per slide (Oren Efraim's golden rule)
- Kill bullet points — max 3 per slide, or use images instead
- Big text — what looks good on laptop doesn't look good projected
Step 4: Content/Media (Do This BEFORE Opening/Closing)
For each slide, check:
- Does it convey exactly ONE idea?
- Is the text big enough to read from the back?
- Are bullet points minimized? (max 3, highlight one at a time)
- Colors: max 5, consistent palette
- Code: monospace, large, syntax highlighted, show LESS not more
- Charts: simplified, noise removed, key data highlighted
- Architecture diagrams: progressive reveal, not everything at once
- Numbers: big and prominent (they're proof, not the story)
Oren Efraim: 10-30 seconds per slide. If you need more time, split the slide.
Step 5: Opening and Closing (LAST — Michal's Rule)
Opening is the LAST thing you write. You need to know all your slides before you know the best hook.
5 Hook Types (Oren Efraim):
| Type | How | Example |
|---|---|---|
| Question | Trigger curiosity | "What happens when you build something for yourself and it becomes a system?" |
| Shared Pain | Empathy hook | "Remember deleting the production database?" |
| Personal Story | Create proximity | "Two years ago I couldn't speak in front of anyone..." |
| Stunning Fact | Challenge assumptions | "We see 1,200 presentations a year. We remember 3%." |
| Paint a Future | Optimism hook | "Imagine a world where every talk you give is remembered." |
Closing: The LAST point MUST callback to the opening. Full circle = closure.
Session 2 output: A complete presentation wireframe — structure, polished slides, opening, closing. Ready to practice.
Session 3: Practice and Polish
Goal: Make the presentation as good as possible. Practice, get feedback, refine.
This is where you take the wireframe from Session 2 and turn it into a performance.
Practice Runs
- Practice out loud — don't just skim in your head (Oren Efraim)
- Memorize only the first 30 sec - 2 min — the opening must be smooth and powerful
- After opening, use slides as anchors — each slide = one message, talk naturally
- Record yourself (video) — watching yourself is the fastest improvement
- Time it: 1 minute = ~150 words
- Have water nearby — drinking = strategic pause
If voice is available (VoiceLayer), use voice_ask for practice runs:
- Speak the opening, get feedback
- Time the full run
- Note where energy drops
Review Checklist
- Premise is clear in the first 2 minutes
- Hook grabs attention (not "Hi, I'm X and today I'll talk about...")
- Each slide = one message
- Energy varies: warm (story) → cool (technical) → warm (insight) → cool (data) → warm (close)
- Closing callbacks to opening
- Timed and within limit
- First 30 sec - 2 min memorized cold
Feedback to Give/Look For
- Where did energy drop?
- Which slide felt too long?
- Was the hook compelling?
- Did the closing feel connected to the opening?
- Audience engagement: would you have stayed?
Present to the Group
Show your opening, get live feedback, iterate. This is the final polish before the real event.
Delivery Rules (Quick Reference)
Voice & Pace
- Speak clearly, vary pace — slow for important, faster for setup
- Use methodical pauses — silence after key points is powerful
- Water = strategic pause when rushing or confused
Body
- Stand, don't sit. Use the stage. Hands visible.
- Face the audience, not the screen
- Eye contact with individuals (3-5 sec each), including back rows
Nerves
- Reframe: excitement, not fear (same physical symptoms)
- Breathing: 5 sec in, 5 sec out (Huberman protocol)
- First 30 seconds are the hardest — it gets easier
- Know your first sentences cold
DON'T
- Don't memorize word-for-word — sounds robotic
- Don't apologize ("This is my first time")
- Don't read slides with your back to the audience
- Don't switch languages mid-sentence (unless technical term)
Environment
- Arrive early, check projector/adapter
- Close all notifications (DND mode)
- Know your mic type (Madonna = belt pack)
- Conferences are always cold — wear sleeves
Audience Engagement
- Raise hands / small jokes work in most cases
- Have a comeback if audience doesn't participate
- The Airplane! trick: Plant a mystery number/fact at the start — someone will ask about it at the end (solves "no questions" problem)
Q&A (Uri Alon)
- Listen to the full question
- Repeat it for the audience
- "I don't know" is valid — follow with "but here's what I think..."
- Long/confused questions: "Let me make sure I understand — are you asking...?"
Voice Integration
When VoiceLayer is available, use it for:
- Practice runs:
voice_ask— speak the opening, get timed feedback - Debrief after presenting:
voice_ask— "How did it go? What would you change?" - Quick notes during review:
voice_speak— capture insights silently (useinsight:prefix) - Reading back the premise:
voice_speak— hear it spoken, check if it sounds natural
Sources
- Oren Efraim — "The Art of Building and Delivering Great Presentations" (TechGym YouTube)
- Michal — Speakers Workshop at Melio (TechGym, Feb 2026)
- Uri Alon — "How to Give a Good Talk" (Weizmann Institute, PubMed)
- Oren Eini — Technical Presentation Delivery Notes (RavenDB/Ayende)
- Ben Orenstein — Speaking for Hackers
- Method Queen — How to Build a Good Lecture (Hebrew)
- Harvard — 10 Tips for Public Speaking
Best Pass Rate
100%
Opus 4.6
Assertions
7
6 models tested
Avg Cost / Run
$0.1241
across models
Fastest (p50)
2.2s
Haiku 4.5
Behavior Evals
Phase 2 baseline — skill quality on ClaudeBehavior Baseline
Adapter Evals
Phase 2C — cross-AI portabilityAdapter Portability
| Assertion | Opus 4.6 | Sonnet 4.6 | Haiku 4.5 | Codex | Gemini 2.5 | Cursor | Consensus |
|---|---|---|---|---|---|---|---|
| starts-with-premise | 6/6 | ||||||
| follows-workshop-phases | 5/6 | ||||||
| includes-opening-and-closing | 5/6 | ||||||
| resists-skipping-premise | 5/6 | ||||||
| asks-about-audience | 6/6 | ||||||
| includes-practice-step | 4/6 | ||||||
| offers-voice-practice | 4/6 |
Token Usage
Cost per Run
| Model | Input Tokens | Output Tokens | Cost / Run | Cost / 1K Runs |
|---|---|---|---|---|
| Opus 4.6 | 5,244 | 6,509 | $0.5668 | $566.80 |
| Sonnet 4.6 | 1,798 | 1,764 | $0.0319 | $31.90 |
| Haiku 4.5 | 2,083 | 1,860 | $0.0028 | $2.80 |
| Codex | 2,990 | 1,843 | $0.0518 | $51.80 |
| Gemini 2.5 | 3,995 | 4,169 | $0.0517 | $51.70 |
| Cursor | 2,270 | 1,929 | $0.0399 | $39.90 |
Response Time (p50)
Response Time (p95)
| Model | p50 | p95 | Overhead |
|---|---|---|---|
| Opus 4.6 | 3.6s | 5.1s | +43% |
| Sonnet 4.6 | 5.6s | 9.6s | +71% |
| Haiku 4.5 | 2.2s | 4.3s | +97% |
| Codex | 6.5s | 10.4s | +59% |
| Gemini 2.5 | 4.4s | 6.9s | +56% |
| Cursor | 6.2s | 8.8s | +42% |
Last evaluated: 2026-03-12 · Data is generated from skill assertions (real cross-model benchmarks coming soon)
Changelog entries are derived from eval runs and skill version updates. Full cascading changelog (Phase 4D) coming soon.
Best Pass Rate
100%
Assertions
7
Models Tested
6
Evals Run
3
- +Initial release to Golems skill library
- +7 assertions across 3 eval scenarios