with one click
tiktok-creator-distiller
// TikTok 博主蒸馏器。输入一个 TikTok 博主账号,采集其热门视频,筛选 TOP 视频,逐条拆解脚本/画面/钩子/节奏,输出账号定位、爆款规律、钩子公式、脚本结构、视觉风格、BGM 策略、选题方向和可复用创作 SOP。适合“帮我蒸馏这个 TK 博主”“分析这个 TikTok 账号为什么火”“把这个博主的爆款打法拆出来”“用 TA 的风格给我的产品写脚本”等请求。
// TikTok 博主蒸馏器。输入一个 TikTok 博主账号,采集其热门视频,筛选 TOP 视频,逐条拆解脚本/画面/钩子/节奏,输出账号定位、爆款规律、钩子公式、脚本结构、视觉风格、BGM 策略、选题方向和可复用创作 SOP。适合“帮我蒸馏这个 TK 博主”“分析这个 TikTok 账号为什么火”“把这个博主的爆款打法拆出来”“用 TA 的风格给我的产品写脚本”等请求。
[HINT] Download the complete skill directory including SKILL.md and all related files
| name | tiktok-creator-distiller |
| description | TikTok 博主蒸馏器。输入一个 TikTok 博主账号,采集其热门视频,筛选 TOP 视频,逐条拆解脚本/画面/钩子/节奏,输出账号定位、爆款规律、钩子公式、脚本结构、视觉风格、BGM 策略、选题方向和可复用创作 SOP。适合“帮我蒸馏这个 TK 博主”“分析这个 TikTok 账号为什么火”“把这个博主的爆款打法拆出来”“用 TA 的风格给我的产品写脚本”等请求。 |
| allowed-tools | Read, Write, Bash |
Follow shared NexTide rules in:
nextide-sharedThis is a NexTide workflow skill for creator-level TikTok research.
It turns one TikTok creator/account into a reusable content operating system:
TikTok creator/account
↓
collect creator videos
↓
select TOP N by likes/views/comments/shares
↓
break down TOP videos
↓
generate creator distillation report
↓
optional: adapt formulas to the user's product/context
Use this skill when the user asks:
Do not use this skill for:
social-data-collectorviral-breakdown-to-video-promptsviral-midform-video-generatorThis is an MVP workflow skill. It does not yet have one monolithic server-side capability.
It orchestrates existing capabilities:
social.tiktok.collect
viral.breakdown.video_prompts
Current limitation:
social.tiktok.collect must support mode=creator in the deployed /api/social-scraper/start workflow.waiting_callback task, do not invent video rows. Wait for the run result/artifacts or ask the user to retry later.Preferred input file:
.nextide/input/tiktok-creator-distill-example.json
Example:
{
"creator": "@quinclips3",
"creatorUrl": "https://www.tiktok.com/@quinclips3",
"platform": "tiktok",
"collectLimit": 20,
"topN": 5,
"sortBy": "likes",
"language": "zh-CN",
"includeVideoBreakdown": true,
"includeScriptTemplates": true,
"productContext": {
"name": "便携榨汁杯",
"sellingPoints": ["便携", "无线", "易清洗"]
}
}
Fields:
creator: TikTok username such as @quinclips3.creatorUrl: optional TikTok account URL.collectLimit: default 20. Keep bounded for first pass.topN: default 5. Do not exceed 8 unless user explicitly confirms cost.sortBy: likes / views / comments / shares / recent.language: report language, default zh-CN.productContext: optional; if present, adapt creator formulas to this product.Create:
.nextide/input/tiktok-creator-collect.json
Shape:
{
"platform": "tiktok",
"mode": "creator",
"targets": ["@quinclips3"],
"limit": 20,
"sortBy": "likes"
}
If the user provided only a URL, put it into targets or urls.
Run:
nextide capability run social.tiktok.collect \
--input .nextide/input/tiktok-creator-collect.json \
--output .nextide/output/tiktok-creator-collect-result.json \
--mode submit \
--wait \
--timeout 900 \
--interval 5
Then export artifacts:
RUN_ID=$(node -e "const r=require('./.nextide/output/tiktok-creator-collect-result.json'); console.log(r.run && r.run.runId)")
nextide run artifacts "$RUN_ID" \
--output-dir .nextide/output/$RUN_ID
Read first:
.nextide/output/$RUN_ID/manifest.json
If the collection run is still waiting_callback, return the runId and tell the user to query later:
nextide run status <run-id>
nextide run result <run-id> --output .nextide/output/<run-id>-result.json
Do not fabricate creator video data.
Find video rows from the result/artifacts. Accept common field names:
video_url / url / post_url / permalink / share_url
caption / description / title
views / play_count / playCount
likes / like_count / diggCount
comments / comment_count / commentCount
shares / share_count / shareCount
posted_at / create_time / createdAt
Rank by sortBy.
Default:
topN = 5
max topN = 8 without explicit confirmation
Write a normalized datatable artifact manually if needed:
.nextide/output/tiktok-creator-distiller/creator-videos.json
Suggested columns:
rank
video_url
caption
views
likes
comments
shares
duration
posted_at
For each selected TOP video, create:
.nextide/input/tiktok-creator-breakdown-01.json
.nextide/input/tiktok-creator-breakdown-02.json
...
Shape:
{
"platform": "tiktok",
"sourcePlatform": "tiktok",
"referenceVideo": "https://www.tiktok.com/@user/video/...",
"sourceTitle": "TOP video 1",
"language": "zh-CN"
}
Run one by one unless the user explicitly asks for parallel processing and accepts cost:
nextide capability run viral.breakdown.video_prompts \
--input .nextide/input/tiktok-creator-breakdown-01.json \
--output .nextide/output/tiktok-creator-breakdown-01-result.json \
--mode submit \
--wait \
--timeout 900 \
--interval 5
Export artifacts for each breakdown run:
RUN_ID=$(node -e "const r=require('./.nextide/output/tiktok-creator-breakdown-01-result.json'); console.log(r.run && r.run.runId)")
nextide run artifacts "$RUN_ID" \
--output-dir .nextide/output/$RUN_ID
If a breakdown returns waiting_callback, save its runId. Continue only when enough finished breakdowns exist. If none finish, return pending run IDs.
Write final report to:
.nextide/output/tiktok-creator-distiller/creator-distillation-report.md
Report structure:
# TikTok 博主蒸馏报告:@creator
## 1. 账号速览
- 博主定位
- 内容主题
- 核心数据概览
- 爆款率
- 更新频率推断
## 2. TOP 视频列表
| Rank | Link | Views | Likes | Comments | Shares | Caption |
## 3. 流量密码:为什么 TA 能爆
- TOP 视频共性规律
- 高互动视频 vs 低互动视频差异
- 分享率最高视频特征
- 完播率推断(必须说明只是基于时长/互动的推断)
## 4. 开头钩子公式
- 公式名
- 原视频钩子原文
- 可套用填空模板
## 5. 脚本结构模型
- 0-3s
- 3-8s
- 8-15s
- 15s+
## 6. 视觉与画面风格
- 拍摄方式
- 构图
- 转场
- 字幕/文字覆盖
- 产品露出方式
## 7. BGM 与声音策略
- BGM 类型
- 人声特点
- 语速/情绪/口头禅
## 8. TOP 视频逐条拆解
For each TOP video:
- 链接
- 数据表现
- 开头钩子
- 脚本结构摘要
- 画面手法
- 一句话归因
## 9. 选题方向 TOP10
## 10. 一条视频的完整创作 SOP
## 11. 用 TA 的风格帮你写脚本
If productContext exists, write 3 sample script directions.
Otherwise ask the user for product/theme/duration.
## 12. 可信度与限制
- 样本数量
- 哪些结论基于真实数据
- 哪些只是推断
- 不鼓励搬运抄袭,只学习方法论
Also write formulas JSON:
.nextide/output/tiktok-creator-distiller/creator-formulas.json
Suggested shape:
{
"creator": "@creator",
"sampleSize": 20,
"analyzedTopN": 5,
"hookFormulas": [],
"scriptStructures": [],
"visualPatterns": [],
"topicDirections": [],
"sop": []
}
Default bounded pass:
collectLimit = 20
topN = 5
Before topN > 8, warn the user:
This will run many video breakdown jobs and may consume more credits. Continue?
Never silently run large batch breakdowns.
unauthorized: ask user to run nextide auth login.insufficient_credits: tell user to recharge NexTide credits.plan_required: tell user the capability requires a paid plan.waiting_callback: return runId and commands to check later.run_not_finished: do not invent results.capability_unavailable: explain which underlying capability is not available.video_url: stop after collection and ask user to provide URLs or wait for normalized artifacts.Final answer to user should include:
Do not paste huge raw JSON. Prefer local artifact paths.