repurpose-tiktok
GitHub基于内容原子生成TikTok视频脚本、图文轮播及拼接/合拍创意。遵循钩子优先结构,优化完播率与互动,适配平台算法规范。
Trigger Scenarios
Install
npx skills add AgriciDaniel/claude-repurpose --skill repurpose-tiktok -g -y
SKILL.md
Frontmatter
{
"name": "repurpose-tiktok",
"license": "MIT",
"metadata": {
"author": "AgriciDaniel",
"version": "1.0.0",
"category": "content"
},
"description": "Generates TikTok content from content atoms: video scripts (15-60s) with hook-first structure and text overlay timestamps, carousel\/photo mode scripts (2-10 slides), and stitch\/duet concepts for trend-riding. Optimizes for completion rate and hook engagement. Sub-skill of the Content Repurposing Engine. Use when user says \"tiktok\", \"tiktok script\", \"tiktok video\", \"tiktok carousel\", \"tiktok stitch\", \"tiktok duet\", \"tiktok photo mode\", or \"repurpose for tiktok\".\n",
"argument-hint": "[url-or-atoms]",
"user-invokable": true
}
TikTok Content Generator
Generate video script, carousel/photo mode, and stitch/duet concept from content atoms.
Inputs
You receive from the orchestrator:
- Content atoms (typed and impact-rated)
- Main argument, target audience, primary topic
- Brand voice profile (5-dimension scores or override flag)
- Flags:
--brief(video script only),--images(generate cover/thumbnail prompts)
References
Load before generating:
references/platform-specs.md-- video specs, carousel format, algorithm weightsreferences/hook-formulas.md-- opening hooks, pattern interruptsreferences/voice-adaptation.md-- TikTok tone rules (fast, casual, raw/authentic)
Platform Rules
| Spec | Value |
|---|---|
| Video length | 15-60s sweet spot (up to 3 min, 10 min for some accounts) |
| Video dimensions | 1080x1920 (9:16) |
| Carousel/Photo mode | 2-10 slides, 1080x1350 (4:5) |
| Algorithm priority | Completion rate #1, then shares, comments, likes |
| Hook window | 0.5-1.5 seconds (determines whether viewer stays or scrolls) |
| Hashtags | 3-5 (mix 1-2 trending + 2-3 niche) |
| Text overlay safe zone | Avoid bottom 20% (TikTok UI overlap) |
| Post types | Video, photo carousel, stitch, duet, LIVE |
| Best time | Tue/Thu/Sat 7-9 PM |
Output 1: Video Script (15-60s)
Write to tiktok/video-script.md.
Timestamp Structure
| Timestamp | Section | Content |
|---|---|---|
| 0-1.5s | Hook | Pattern interrupt, bold claim, or unexpected question. This single moment determines everything. |
| 1.5-5s | Context | Quick setup: who you are OR why this matters. Keep it raw, not polished. |
| 5-25s | Value | Core insight from 2-3 highest-impact atoms. Jump cuts every 2-3 seconds. Each sentence must earn the next. |
| 25-45s | Proof | Stat, result, example, or before/after. Show, don't just tell. |
| 45-55s | CTA | Follow, comment, stitch, or "link in bio." Only ONE ask. |
| 55-60s | Loop hook (recommended) | Final line connects back to the opening -- encourages rewatch (boosts completion rate significantly). |
Video Script Rules
- Every second must justify the next. If a viewer can skip 3 seconds and miss nothing, cut those 3 seconds.
- Jump cuts every 2-3 seconds minimum. No static talking-head shots longer than 3s.
- Write the script as spoken words, not written prose. Fragments. Pauses. Emphasis marks.
- Completion rate is king. A 30s video with 90% completion massively outperforms a 60s video with 40%.
- The hook must work with sound OFF (text overlay carries it) and sound ON (voice carries it).
Text Overlay Plan
For each timestamp section, specify:
| Timestamp | On-Screen Text | Position | Animation |
|---|---|---|---|
| 0-1.5s | [Hook phrase, 3-6 words] | Top-center | Pop-in or zoom |
| 1.5-5s | [Context keyword] | Center | Typewriter |
| 5-25s | [Key phrases per jump cut] | Top or center | Cut-sync |
| 25-45s | [Stat or result number] | Center-bold | Scale-up |
| 45-55s | [CTA text] | Center | Pulse |
Critical: Avoid bottom 20% of frame for all text overlays -- TikTok's UI (caption, buttons, progress bar) covers this area.
Audio Direction
- Option A: Voiceover -- The script IS the voiceover. Write it conversationally with natural pauses.
- Option B: Trending sound -- Note: "Use trending audio for discoverability. Voice as text overlay only."
- Option C: Original audio -- Direct-to-camera delivery. Mark emphasis and pace changes.
- Always recommend trending sound research: "Check TikTok's Creative Center for current trending sounds in [niche]."
Tone Rules
| Do | Do NOT |
|---|---|
| Use casual, conversational language | Use corporate or formal phrasing |
| Start mid-thought ("So I just found out...") | Start with introductions ("Hi, I'm...") |
| Use trending phrases and patterns | Use dated slang or forced memes |
| Speak in short, punchy fragments | Write in complete, polished sentences |
| Be raw and authentic | Over-produce or over-edit the feel |
| Use "you" and "I" naturally | Lecture or teach from above |
| Show genuine reactions and opinions | Be neutral or hedging |
Output 2: Carousel / Photo Mode (2-10 Slides)
Write to tiktok/carousel.md.
Slide Structure
| Slide | Purpose | Content |
|---|---|---|
| Slide 1 | Stop the scroll | Hook headline (5-8 words), bold visual, "Swipe" implied by format |
| Slides 2-5 | Core value | One idea per slide from highest-impact atoms. Text-forward, minimal graphics. |
| Slides 6-8 | Proof + examples | Stats, comparisons, or actionable steps |
| Final slide | CTA | Follow + save prompt, or teaser for related content |
Carousel Rules
- Dimensions: 1080x1350 (4:5) for all slides
- TikTok carousels are text-forward -- prioritize readability over visual complexity
- One idea per slide, large readable text
- Consistent visual style across slides
- TikTok carousel captions are shorter than Instagram -- keep to 150-300 chars
- Photo mode is a rising format with lower competition than video (2026)
Caption for Carousel
- 150-300 characters (shorter than Instagram)
- Hook in first line
- 3-5 hashtags (mix trending + niche)
- End with engagement prompt ("Save this for later" or "Which one surprised you?")
Output 3: Stitch/Duet Concept
Write to tiktok/stitch-duet.md.
Concept Structure
| Field | Content |
|---|---|
| Trending angle | Describe the type of trending content to respond to (not a specific video) |
| Your hook | Opening line for the stitch/duet (0.5-1.5s) |
| Response script | 15-30s script with your angle from the atoms |
| Why it works | 1-2 sentences on why stitch/duet boosts reach (algorithm loves participatory content: 20-30% more reach) |
Stitch vs Duet
- Stitch: Use the first 1-5s of another video, then cut to your take. Best for "reacting to" or "adding to" a trending topic.
- Duet: Side-by-side with another video. Best for comparisons, demonstrations, or visual responses.
Concept Rules
- Suggest the TYPE of trending content to stitch/duet with, not a specific creator or video
- The response portion must deliver standalone value from the atoms
- Frame as contributing to a conversation, not stealing content
- Keep the response to 15-30 seconds
Image Prompts
If --images flag is set, suggest:
- Video thumbnail: 9:16 (1080x1920) eye-catching cover frame concept
- Carousel slides: 4:5 (1080x1350) visual concepts for key slides
- Include /banana prompt descriptions
Output Format
Each file should include content plus production notes:
# TikTok Video Script: [Topic]
## Hook (0-1.5s)
**Say:** "[spoken text]"
**Text overlay:** [on-screen text] | Position: [top/center] | Animation: [type]
**Visual:** [what the viewer sees]
## Context (1.5-5s)
...
---
**Production notes:** [audio direction, posting time, hashtag suggestions, trending sound research]
Brief Mode
If --brief is active, produce only:
- 1 video script (15-60s)
- Skip carousel and stitch/duet concept
Error Handling
| Condition | Action |
|---|---|
| Fewer than 3 atoms | Write a shorter script (15-30s); focus on one insight |
No stat or casestudy atoms |
Skip the Proof section; extend Value section |
brief_mode is true |
Generate video script only; skip carousel and stitch/duet |
| Content is too academic/formal | Increase casual rewording; add "imagine this..." framing |
| Topic doesn't suit video | Prioritize carousel/photo mode over video script |
| Content is purely promotional | Reframe around the insight; strip product mentions |
Version History
- 669187e Current 2026-07-05 08:56


