ppt-classify — PPT Type Classification + Chapter Skeleton + Argumentation Routing
When to Trigger
Applicable Scenarios:
- The user describes a PPT task ("Help me create a PPT about xxx") but hasn't decided on the argumentation approach yet
- Before starting to make the PPT, the user wants to clarify the overall structure first
- During team discussions on PPT direction, it's necessary to align on "what type of presentation this is" + "how many chapters there will be in total"
Inapplicable Scenarios:
- The PPT type is already clear and chapters are finalized → Directly use the corresponding argumentation skill to dive into each chapter
- Pure data reports / pure text documents → This is not a PPT-related issue; switch formats
Core Insight: Classification + Skeleton = Step 0
The first mistake in making a PPT is not choosing the wrong template, but not clarifying what type of PPT you're creating and not thinking about the overall skeleton.
The argumentation approaches for the four PPT types are completely different, and their corresponding chapter skeletons are also entirely different:
| Type | Core Objective | Argumentation Approach | Typical Skeleton Form | Typical Scenarios |
|---|
| Pitch / Proposal | Persuade the audience to accept a certain judgment | Present a viewpoint → Defend it | Pain Point → Solution → Evidence → Path → CTA | Financing / Project Initiation / Proposal / Conclusion Reporting |
| Research / Investigation | Guide the audience through an investigation | Raise a question → Design a path → Derive a conclusion | Paradox → Argumentation of the Research Subject → Multiple Verification Paths → Synthesis | Industry Analysis / Competitor Research / Investment Research |
| Teaching / How-to | Teach the audience to do something | Clarify target competence → Break down steps | Objective → Prerequisites → Step-by-Step → Verification → Extension | Training / Tutorial / Methodology Popularization |
| Narrative / Storytelling | Tell a story | Protagonist + Conflict + Twist | Opening → Conflict → Confrontation → Breakthrough → Aftermath | Project Review / Brand Story / Case Sharing |
Using a pitch-style approach for a research PPT → You'll come across as an advocate.
Using a research-style skeleton for a pitch PPT → The audience won't know what you want them to do after listening.
Using a narrative-style skeleton for a teaching PPT → The audience is moved by the story but learns nothing.
Diagnostic Process
Ask questions layer by layer (in order; stop when the previous question hits):
Q1: Does this PPT require the audience to make a specific decision?
- Yes → Most likely a Pitch (Financing, product initiation, proposal approval are all decision-driven)
- Further confirmation: Is the conclusion already finalized? Are you defending a judgment? → Confirm Pitch
- If the conclusion is not finalized and you just "want the audience to participate in the decision" → May be Research
- No → Proceed to Q2
Q2: Does the audience need to be able to do a specific task after listening to the PPT?
- Yes (Able to independently reproduce an operation or apply a method) → Teaching
- No → Proceed to Q3
Q3: Is the core driving force "specific event process" or "abstract argumentation"?
- Event process (Has protagonist, timeline, emotional arc) → Narrative
- Abstract argumentation (Has framework, evidence chain, dimensional analysis) → Research
Q4: Edge Cases
- Hybrid type: Mark the main type + secondary type. The main type determines the argumentation skill + skeleton form, while the secondary type determines auxiliary techniques
- User is uncertain: List the two most similar types, describe the argumentation consequences and skeleton differences for each, and let the user choose
- Should not make a PPT at all: Such as pure data query, pure FAQ → Suggest switching formats
Output Format · v0.2.0 New Addition: Chapter Skeleton Suggestions
After completing the diagnosis, output the following structure (must include four sections):
【PPT 类型判定】
主类型：Research
次类型：（无 / 或注明）
判定依据：用户想带观众走一遍"NVIDIA 护城河是否削弱"的调查，
结论不预设，观点在多条路径汇合后产生。
【推荐立论 skill】
主路径：ppt-research-setup（研究型三段式 + specificity 诊断 +
本 skill 输出的章节骨架作为输入）
辅助：完成后可用 ppt-narrative-review 做张力审稿
【章节骨架建议】
研究型典型骨架（已针对"NVIDIA 护城河"personalize）：
章 1 · 反共识悖论 | 翻倍悖论：$500B → $1T · 21 个月
章 2 · 被研究对象论证 | Jensen 的四步推理链（需求 / Token / TCO / 资金）
章 3 · 检验路径 1 | 需求真爆炸？三行业 Token 落地 + 物理极限
章 4 · 检验路径 2 | Token 经济 + NVIDIA 身份换皮
章 5 · 检验路径 3 | TCO 护城河 + 中国替代链威胁
章 6 · 综合判决 | 四步结论 + 3 个重估信号
建议章节数：5-7 章（研究型的适配范围）
建议单集长度：如 5 min 视频配套，每章约 30-60 秒；如 20 min 深度，可扩到每章 2-3 分钟
【下一步】
运行 ppt-research-setup，把上面的章节骨架传入，让它把每章填成
具体的推理链 + 证据锚点。
Canonical Skeletons for Four Types (Reference)
This is the canonical skeleton used for internal lookup in the skill. Replace chapter titles with specific topics when personalizing:
Pitch / Proposal (5-7 chapters)
- Pain Point / Current Situation (Why the audience should care)
- Solution (What we propose)
- Why Us (Differentiation / Team / Assets)
- Traction / Evidence (Users / Orders / Experimental Data)
- Business Model / Path (Where the money comes from / How to scale)
- (Optional) Risk Mitigation
- CTA / Ask (What the audience should do when leaving the meeting)
Research / Investigation (5-7 chapters)
- Counterconsensus Paradox (Research motivation: Why this question is not obvious)
- Argumentation of the Research Subject (Explain it fully first; don't rush to refute)
3-5. Verification Path 1/2/3 (Each is independent, verifiable, and can produce evidence)
- Synthetic Judgment + Monitoring Framework (Conclusion + Future re-review signals)
Teaching / How-to (4-6 chapters)
- What You Can Do After Learning (Target competence / Acceptance criteria)
- Prerequisites (What basics you need / What tools to prepare)
3-5. Step-by-step (Can be split into 2-4 chapters based on complexity)
- Common Pitfalls + How to Verify + Extended Learning
Narrative / Storytelling (5 chapters · Three-Act Structure)
- Opening: Protagonist + World Setting
- Twist 1: Conflict / Trigger Event
- Middle Act: Escalating Confrontation / Cost
- Twist 2: Breakthrough / Epiphany / Critical Decision
- Ending: New Normal + Aftermath / Inspiration
Personalization Rules
Do not directly output the original canonical skeleton above. For the user's specific topic:
- Identify key concepts in the topic (e.g., "NVIDIA Moat", "Claude Code Tutorial")
- Replace the abstract slots in the canonical skeleton with specific chapter titles
- Check if the number of chapters matches the user's expected duration
- If it's a hybrid type (Main X + Secondary Y), embed 1-2 chapters of the secondary type in the skeleton (Example: Main Research + Secondary Narrative → Add a "Key Case" sub-chapter in one of the verification paths)
Quick Reference: Mapping of Four Common Scenarios
| User Description | Classification | Skeleton Example |
|---|
| "Help me create a financing PPT" | Pitch | Pain Point → Solution → Traction → Funding → CTA |
| "Help me create a research PPT about xx industry" | Research | Paradox → Subject Argumentation → 3 Verifications → Synthesis |
| "Help me create a usage tutorial PPT for xx product" | Teaching | Objective → Prerequisites → 3 Steps → Verification |
| "Help me create a project review PPT" | Narrative | Opening → Conflict → Confrontation → Breakthrough → Aftermath |
| "Help me create a PPT to report this quarter's KPI" | Main Pitch + Secondary Narrative | Main Pitch skeleton + Hook with a specific case in the opening |
| "Help me create an analysis of NVIDIA's moat" | Research | (See the example in the Output Format section above) |
| "Help me create a Claude Code tutorial" | Teaching | What you can do → Install CLI → Basic Commands → Advanced → Verification |
| "Help me create a proposal PPT for clients" | Main Pitch + Secondary Research | Pitch skeleton + Use research-style evidence organization in the argument chapter |
Handling Hybrid Skeletons
The main type determines the skeleton form. The secondary type is added as an embellishment in one chapter without changing the overall skeleton:
- Main Research + Secondary Narrative: Insert a case story in a subsection of "Verification Path X"
- Main Pitch + Secondary Research: Organize evidence using a research-style framework inside the argument chapter (Chapter 4)
- Main Teaching + Secondary Narrative: Change the opening pain point chapter to "A Pitfall Story"
- Main Narrative + Secondary Research: Use counterintuitive data as the trigger for turning points (Chapters 2/4)
Claude Integration
When the user's request triggers this skill:
- If the user's description is clear enough: Directly diagnose using the Q1-Q3 chain, and provide the classification + personalized skeleton
- If information is insufficient: Use to inquire layer by layer. Must ask three items: Topic · Materials / First-hand Sources · Target Audience · (Optional) Duration Expectation
- Must personalize when outputting the skeleton: The canonical skeleton is a draft; the final product must be tailored to the specific topic
- Do not handle argumentation: This skill only provides the type + skeleton + name of the next skill. The specific reasoning chain for each chapter is filled by downstream skills like
Contract with Downstream Skills
Downstream skills like
will take the skeleton output by this skill as
input:
- outputs chapter names + chapter roles · strategic level
- fills each chapter with Section 1 (Specific wording of counterconsensus paradox) + Section 2 (Reasoning chain) + Section 3 (Specific anchor points for verification paths) · tactical level
The two layers of responsibilities do not overlap: classify manages "what chapters this presentation has", while research-setup manages "what logical chain each chapter specifically covers"
Design Origin
This skill is refined from Experiment A of EP10《AI Makes PPT》. Experiment A found that applying pitch-style argumentation skills to research-type PPTs would turn researchers into advocates — classification is the true Step 0 in making PPTs.
v0.2.0 Changes: Before the official recording of EP10, it was found that the original v0.1.0 only provided type labels without skeletons, leading to a gap between users "knowing what type of PPT it is" and "knowing how many chapters there should be". v0.2.0 moves the skeleton suggestion to the classify step, allowing Step 0 to truly serve as "overall positioning" — providing strategy + skeleton, while downstream research-setup only focuses on filling in the details of each chapter.
For details:
notes/experiment-a-ceo-review.md
and 《Classification of PPT Presentations and Application in Work Scenarios》.