# Video Analysis

> Analyze Douyin and Xiaohongshu content cases for reusable short-video strategy. Use when the user asks for "video analysis", "va", content teardown, case拆解, 爆款分析, 广告案例分析, 对标账号分析, 同赛道案例分析, 抖音案例, 小红书案例, or wants to judge content value, extract reusable structures, and adapt examples for accounts including 苏苏suri, 年年nroa, 科技御姐jojo, or 数码喵小葵.

- Skill: `vanyaan77-lgtm/video-analysis` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add vanyaan77-lgtm/video-analysis`
- Raw SKILL.md: https://api.skillmd.com/api/skills/vanyaan77-lgtm/video-analysis/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: vanyaan77-lgtm (https://skillmd.com/u/vanyaan77-lgtm)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/vanyaan77-lgtm/video-analysis

---


# Video Analysis

## Core Rule

Use this skill only for Douyin and Xiaohongshu content cases. Analyze any useful reference case, including viral posts, ads, competitor posts, same-category examples, benchmark accounts, and partial materials.

Do not only summarize what the video says. Prioritize:

- Why people watch.
- Why people interact.
- Which structures are worth borrowing.
- Which parts are not worth copying.
- How to adapt the case to the user's accounts.

Always produce structured, execution-oriented conclusions. If information is incomplete, analyze what is available and ask for the most useful missing materials instead of ending the analysis.

Exception: if the user provides only a video link and the actual video画面 cannot be accessed or seen, do not continue into content structure teardown, pacing/camera analysis, reusable templates, or account adaptation. Output only the information-completeness rating, the blocker, and the exact material needed next. Do not make title-only guesses for video拆解.

## Workflow

### 1. Identify Material Completeness

Classify the current material as one or more of:

- A. Complete video file
- B. Video link
- C. Screen recording
- D. Video screenshots
- E. Title + cover
- F. Title + data
- G. Title + comment section
- H. Partial subtitles

Output:

```markdown
【信息完整度评级】
评级：高 / 中 / 低
当前可以分析：
- ...
当前无法分析：
- ...
建议补充：
- 视频文件 / 录屏 / 字幕 / 评论区 / 链接
```

Use the rating this way:

- 高: Video or screen recording plus enough title/data/comment context to assess structure, pacing, visuals, and audience reaction.
- 中: Enough text, screenshots, link, subtitles, or metrics to infer part of the structure, but visual pacing or comments may be missing.
- 低: Only title, cover, sparse data, or fragments; still infer hook, positioning, and possible value, while clearly marking uncertainty.

If the material is only an inaccessible link, stop after this section and request a recording, visible-page capture, screenshots, subtitles, or comments. Do not run steps 2-8 unless there is visible video evidence or user-provided content beyond the blocked link.

### 2. Extract Basic Information

Create a basic information table. Use "未提供" or "无法判断" when unavailable.

Required fields:

- 标题
- 平台
- 时长
- 发布时间
- 点赞
- 收藏
- 评论
- 转发
- 粉丝量

Output heading: `【基础信息表】`

### 3. Identify Content Type

Choose the closest type:

- 广告
- 种草
- 测评
- 横测
- 开箱
- 口播
- Vlog
- 情绪分享
- 知识科普
- 剧情
- 探店
- 其他

Explain the evidence briefly, such as product露出, CTA, personal experience, comparison structure, plot setup, emotional confession, or educational framing.

### 4. Deconstruct Content Structure

Output heading: `【内容结构拆解】`

#### Opening Hook

Identify the hook type:

- 反差
- 悬念
- 痛点
- 情绪
- 争议
- 身份背书
- 数据震惊

Explain why it may create retention. Tie the reason to audience psychology, unmet need, curiosity gap, conflict, identity relevance, or practical value.

#### Chapter Structure

Break the content into segments. If exact timestamps are unavailable, use approximate sequence labels.

For each segment include:

- 时间轴
- 内容目的
- 情绪变化

Use a compact table when possible.

#### Pacing

Assess:

- 信息密度
- 转折频率
- 刺激点
- 停留点

Classify as `快节奏`, `中节奏`, or `慢节奏`, and explain why.

#### On-screen Text

Analyze flower text/subtitle/card text if visible or inferable:

- 出现频率
- 作用: 强调 / 补充 / 反转 / 情绪放大
- 哪些文案可学，哪些不建议照搬

#### Camera And Visuals

If visuals are available, analyze:

- 机位
- 运镜
- 切镜频率
- 素材类型
- B-roll 使用情况

If visuals are unavailable, explicitly say visual analysis is limited and infer only from provided screenshots, cover, or subtitles.

### 5. Comment Section Insights

When comments are available, extract:

- 高赞评论
- 争议点
- 用户关注点
- 用户吐槽点
- 用户真实需求

Output heading: `【评论区洞察】`

If comments are unavailable, state what audience reaction cannot be verified and suggest supplementing comment screenshots or copied comments.

### 6. Summarize Case Value

Output heading: `【案例价值总结】`

Include:

1. 内容目标: classify as 广告 / 种草 / 测评 / 情绪内容 / 日常分享 / 品牌宣传 / 转化内容.
2. 核心亮点: evaluate 标题 / 开头钩子 / 节奏 / 花字 / 镜头 / 情绪设计 / 转化设计.
3. 可借鉴程度:
   - `★★★★★` 适合直接借鉴
   - `★★★★` 可借鉴大部分结构
   - `★★★` 可借鉴部分思路
   - `★★` 参考价值有限
   - `★` 不建议借鉴
4. 借鉴理由: explain concretely why it is or is not worth learning.

### 7. Extract Reusable Templates

Output ready-to-use templates under these headings:

- `【标题模板】`
- `【开头钩子模板】`
- `【内容结构模板】`
- `【评论引导模板】`
- `【转化模板】`

Templates must be directly fillable, using placeholders such as `[人群]`, `[痛点]`, `[产品/场景]`, `[反差结果]`, `[具体证据]`, `[行动引导]`.

### 8. Adapt For The User's Accounts

Output a separate subsection for each account:

- 苏苏suri
- 年年nroa
- 科技御姐jojo
- 数码喵小葵

For each account include:

1. 是否适合改编
2. 推荐改编方向
3. 推荐内容形式: 日常脚本 / 广告脚本 / 抖音内容 / 小红书内容
4. 推荐标题方向
5. 推荐开头钩子

Keep recommendations specific to the case. Avoid generic advice like "加强互动" unless paired with an exact comment prompt, scene, title, or script angle.

## Output Style

- Write in Chinese by default unless the user asks otherwise.
- Use clear headings in the exact bracket style where specified.
- Mark assumptions and uncertainty when materials are incomplete.
- Prefer concise tables and actionable bullets.
- End with the highest-value next action, such as which missing material would most improve the analysis or which account should adapt the case first.