メニュー
Automate TikTok marketing across multiple apps using Dropspace for publishing/analytics. Use when setting up TikTok content automation, creating slideshow posts, analyzing post performance, optimizing marketing funnels, or when user mentions TikTok growth, slideshows, or social media marketing for any app.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Generate images and videos via Fal.ai API. Use when generating AI images (FLUX, nano-banana-2), AI videos (Veo 3.1), or any Fal.ai model. Supports queue-based async generation with polling, reference image editing, and file download. Also use when the content pipeline needs direct image/video generation bypassing Dropspace rate limits.
Content clipping engine — process videos and images from Google Drive into short-form content. Video clips via audio peak detection + FFmpeg. Image slideshows via Drive folder selection + canvas overlays. Use when user asks to clip videos, create slideshows from photos, process footage, or create short-form content from existing media.
Self-improving content pipeline for Dropspace. Analyzes post performance, generates new content across 6 platforms, schedules publishing, and compounds strategy over time. Run nightly for best results. Use when asked to run the content engine, generate social posts, or manage the content pipeline.
Extract short clips from DJ set recordings and schedule to TikTok/Instagram via Dropspace. Uses ffmpeg for audio analysis and video cropping. Use when asked to clip DJ sets, extract video highlights, or create short-form music content.
Turn event photos from Google Drive into TikTok/Instagram slideshows via Dropspace. Face-aware text positioning, community voice captions. Use when asked to create slideshows from event photos or manage photo-based social content.
Create and publish social media launches via the Dropspace API. Use when user wants to post to social media, create launches, manage personas, track post analytics, schedule content, or anything Dropspace-related. Supports 10 platforms (Twitter, LinkedIn, Reddit, Instagram, TikTok, Facebook, YouTube, Substack, Product Hunt, Hacker News) with per-post analytics and AI content generation.
| name | tiktok |
| description | Automate TikTok marketing across multiple apps using Dropspace for publishing/analytics. Use when setting up TikTok content automation, creating slideshow posts, analyzing post performance, optimizing marketing funnels, or when user mentions TikTok growth, slideshows, or social media marketing for any app. |
Automate TikTok slideshow marketing: generate → overlay → publish via Dropspace → track → iterate.
All platforms share a common engine at ~/markus/. No per-platform scripts — everything runs through shared engines with platform config from platforms.js.
Data: ~/markus/apps/{app}/tiktok/ (strategy.json, posts.json, failures.json, pending-batches.json, posts/)
App config: ~/markus/apps/{app}/app.json (single source of truth for app identity, integrations, platform settings)
~/markus/)# Create a post (auto-picks from queue)
node engines/create-visual-post-engine.js --app <APP> --platform tiktok --next --schedule "2026-03-04T12:00:00Z"
# Pre-generate images via OpenAI Batch API (50% cheaper)
# Self-improve (analytics + strategy optimization)
node engines/self-improve-engine.js --app <APP> --platform tiktok --days 14
# Validate setup
# Add posts to queue (pipe JSON to stdin)
echo '{"posts":[...],"notes":"..."}' | node scripts/add-posts.js --app <APP> --platform tiktok
story-slideshow (default) — 5 content slides + engine CTA (1024×1536 portrait).SELF_ONLY — posts land as drafts in your TikTok inbox. He adds trending sound, changes privacy to public.auto_add_music: true as fallback. Manual trending sound = ~100x more views.| Slide | Purpose | Text Style |
|---|---|---|
| 1 | HOOK — stop the scroll | Full hook text, relatable problem |
| 2 | PROBLEM — amplify pain | Build tension |
| 3 | DISCOVERY — turning point | "So I tried this..." |
| 4 | TRANSFORMATION 1 — first result | "Wait... this actually works?" |
| 5 | TRANSFORMATION 2 — escalate | "Okay I'm obsessed" |
Note: Slide 6 (CTA) is engine-appended. The LLM generates 5 content slides only.
Sizing:
Positioning:
Content:
[hook matching slide 1] 😭 [2-3 sentences of relatable struggle].
So I found [APP NAME] that [what it does in one sentence] —
you just [simple action] and it [result].
I tried [thing 1] and [thing 2] and honestly?? [emotional reaction].
[funny/relatable closer]
#hashtag1 #hashtag2 #hashtag3 #hashtag4 #fyp
Rules:
Tier 1: Person + Conflict → Result (consistently 50K+)
Tier 2: Relatable Pain → Discovery
Tier 3: Curiosity / Challenge
What DOESN'T work: Feature lists, self-focused complaints without conflict, fear/insecurity hooks.
When generating posts for the queue via add-posts.js, the blueprint structure depends on the format:
story-slideshow uses sceneAnchor + slideMoods (locked architecture — see formats.js and formats.js story-slideshow entry):
{
"format": "story-slideshow",
"text": "Hook text",
"slideTexts": ["Slide 1 text", "Slide 2", "Slide 3", "Slide 4", "Slide 5"],
"sceneAnchor": "Detailed scene (150-300 words): room, furniture, person, camera. Shared across all slides.",
"slideMoods": ["mood delta 1 (30-60 words)", "mood 2", "mood 3", "mood 4", "mood 5"],
"caption": "Storytelling caption with #hashtags"
}
{
"text": "Hook text",
"videoPrompt": "6-second video description...",
"caption": "Caption"
}
Engine auto-appends CTA slide for formats with ctaSlide: true.
Note: The pipeline now uses engagement rate as the primary optimization signal. Raw view thresholds are secondary guidance.
| Views | Conversions | Diagnosis | Action |
|---|---|---|---|
| 🟢 High | 🟢 High | SCALE | 3 variations of winning hook |
| 🟢 High | 🔴 Low | FIX CTA | Rotate CTAs, check landing page |
| 🔴 Low | 🟢 High | FIX HOOKS | Test different hooks, keep CTA |
| 🔴 Low | 🔴 Low | FULL RESET | New format, new audience angle |
First-touch attribution (if configured) captures referral source:
| Mistake | Fix |
|---|---|
| Landscape images (1536x1024) | Portrait always (1024x1536) |
| Font too small (5%) | 6.5% of width minimum |
| Text at top/bottom of image | Position at 28% from top |
| Different scenes per slide | Lock architecture, only change style |
| Labels instead of reactions | "Wait this is nice??" not "Modern style" |
| gpt-image-1 instead of 1.5 | Always gpt-image-1.5 |
| Generic image prompts | Be obsessively specific |