Deconstruct a piece of content from the audience's perspective—not how it was written (the creator's perspective), but what the audience got after watching it, where they got it, and what work each sentence is doing.
Trigger method: /cuihuo-audience-analysis, "deconstruct this one," "audience-perspective analysis," "what value does this provide to the audience," "deconstruct sentence by sentence"
Core framework: five-dimensional value model (MECE)
The value any piece of content provides to the audience falls into, and only into, these five dimensions:
| Dimension | Definition | Test method |
|---|---|---|
| Cognitive value | The audience learns something they did not know before—a fact, concept, causal relationship, or framework | After watching, they can say "so that's how it is" |
| Emotional value | The audience's feelings change—resonance, catharsis, fear, urgency, satisfaction | After watching, there is a physical reaction: heart rate speeds up, they want to cry, they want to repost and curse someone out |
| Identity value | The audience's understanding of who they are changes—permission ("I can be unkind"), a new identity ("I am a clear-headed person"), belonging ("I belong to this camp") | After watching, they will use a new word to describe themselves, or allow themselves to do something they previously did not dare to do |
| Action value | The audience knows what to do next—specific operations, scripts, steps, checklists | After watching, they can immediately go do one clear thing |
| Social value | The audience gets something they can use in relationships—talking points, a stance to repost, material to teach others | After watching, their first reaction is "who should I send this to" or "I can bring this up next time I chat" |
Mutual-exclusivity check: the five dimensions change, respectively, what you know, what you feel, who you think you are, what you do, and what you display to others.
A piece of content, a paragraph, or a sentence can provide value on multiple dimensions at the same time. When labeling, do not choose a single category; instead mark all hit dimensions and their intensity (strong/medium/weak/none).
Three-layer analysis
Layer one: overall layer
For the entire piece of content (topic + full text), answer:
- What the topic is — Say clearly in one sentence what this piece of content is about. Do not use the creator's title; use the audience's language: "This video is telling me..."
- Topic promise — The expectation the title and opening give the audience: what type of gain does it promise you? (loss-avoidance insider knowledge / operation manual / identity declaration / emotional outlet / ...)
- Main value dimensions — Which 1-2 of the five dimensions are the main ones for this piece of content? Say clearly in one sentence: this piece of content is essentially not teaching you something, it is permitting you to become another person (identity main) / this piece of content is essentially an operation manual (action main) / ...
- Hidden tension hit — What is that thing that already exists in the audience's mind but has not been spoken aloud. It is not the topic the creator chose to talk about; it is the feeling that was hit in the audience's mind.
- Content Type Judgment — Scattershot type (multiple parallel arguments spanning multiple topics) or focused type (one concept/path hammered from start to finish)
Second Layer: Node Layer
Break the content into logical nodes (hook / each argument or section / closing CTA), and annotate each node across five dimensions:
### 节点 N：{节点标题}
- 认知：{强/中/弱/无} — {具体说明：观众在这里知道了什么}
- 情绪：{强/中/弱/无} — {具体说明：观众在这里感受到了什么变化}
- 身份：{强/中/弱/无} — {具体说明：观众对自己的认知怎么变了}
- 行动：{强/中/弱/无} — {具体说明：观众拿到了什么可执行的指令}
- 社交：{强/中/弱/无} — {具体说明：观众拿到了什么社交素材}
主力价值：{1-2 个维度} + {一句话说这个节点在干什么活}
After the node layer analysis is complete, add a summary of the relationships between nodes:
- Whether the main value of each node is consistent (all doing identity permission vs. each node doing a different job)
- Whether there are outlier nodes (ones different in nature from the other nodes, for example a pure action-operation node sandwiched among five identity permission nodes)
- The emotional curve of the node arrangement: where the emotional peak is, where the information-dense zone is, where the action landing point is
Third Layer: Sentence-by-Sentence Layer
Annotate which of the five dimensions each sentence serves. Format:
| 原句 | 认知 | 情绪 | 身份 | 行动 | 社交 | 说明 |
|------|:----:|:----:|:----:|:----:|:----:|------|
| {原句} | {·/★} | {·/★} | {·/★} | {·/★} | {·/★} | {这句话在干什么} |
★ = hit, · = miss. A sentence can have multiple ★.
The sentence-by-sentence layer does not need to be done every time — use it only when you need to finely break down a single node, or need to find "which sentence is the most powerful sentence in this video." By default, only do the overall layer + node layer. Enter the third layer only when the user says "break it down sentence by sentence."
Input
Accepts the following inputs:
- The complete copy of one video/image-text post (text)
- A file path (read then analyze)
- A crawler data folder path (automatically find 原视频脚本.md / content.md)
If the input is a folder path, find the content file in the following priority order: 原视频脚本.md > content.md > *.md
Output Rules
- Do not evaluate whether the content is good or bad. This is not a content audit; it is value decomposition.
- Do not use creator terminology (hook templates, CTA design, shell reuse) — those are words from the creator's perspective. Use language the audience can feel.
- The five-dimension annotations must have specific explanations; you cannot just write "strong" without saying why it is strong.
- The overall layer's "latent tension that was stepped on" must be written in the audience's language, not the creator's topic description. Difference: the creator says "girls need to be ruthless," but what the audience feels is "I always get the short end of the stick in relationships because I'm too soft."
- Each node in the node layer must point out its main value dimension; they cannot be distributed evenly. Even distribution means the analysis was not clear.
- After the analysis is complete, write a one-sentence summary at the end: what this content is essentially doing (similar to "this video is not teaching you something, it is permitting you to become another person").
Relationship with Other Skills
- This skill only does decomposition, not transfer. After decomposition is complete, if the user wants to do topic transfer, go through the topic transfer process (to be built).
- Difference from Material Deep Reverse Engineering (/business-素材深度逆向): Material Deep Reverse Engineering decomposes the creator's techniques and level-of-consciousness transfer, while this skill decomposes the value the audience actually receives. The two are complementary and do not conflict.
- Difference from Influencer Analysis (/influencer-creator-analysis): Influencer Analysis is account-level systematic analysis, while this skill is single-content-level decomposition.
Language
- If the user uses Chinese, use Chinese; if they use English, use English
- Chinese replies follow the 《中文文案排版指北》