# Absorbed skill: shortform-growth-automation

---
name: shortform-growth-automation
description: "Design and operate semi-automatic pipelines for YouTube Shorts, TikTok, Instagram Reels, and similar shortform growth: niche strategy, test batches, human review loops, monetization constraints, and analytics feedback."
version: 1.0.0
author: Hermes Agent
license: MIT
metadata:
  hermes:
    tags: [shorts, tiktok, youtube-shorts, instagram-reels, growth, social-media, content-automation, monetization]
    related_skills: [youtube-content, google-workspace, github-repo-management]
---

# Shortform Growth Automation

Use this skill when building, reviewing, or improving systems that generate shortform social-media videos, especially semi-automated pipelines for YouTube Shorts, TikTok, Instagram Reels, Facebook Reels, or repurposed longform content.

## Core principle

Treat the system as a **growth engine**, not merely a video generator. A working renderer is only useful if it produces content with strong hooks, retention, identity, and feedback loops.

## Recommended workflow

1. **Define content pillars**
   - Start with 2-4 clear pillars, not a random topic list.
   - For each pillar define: audience, promise, tone, no-go topics, hook patterns, CTA patterns, visual rules.
   - Prefer identity-rich pillars such as `AI as a servant`, `Faith and focus`, `Life systems for overloaded adults` over generic advice like “health tips”.

2. **Build a test batch before automation**
   - Create a small batch, e.g. 9 ideas: 3 per pillar.
   - Include a mix of short growth candidates and longer monetization candidates.
   - Each idea should include: id, pillar, title, platform target, duration, hook, script outline, visual concept, caption, CTA, quality score, and risk notes.

3. **Score scripts before rendering**
   - Hook in the first 1-2 seconds.
   - Clear emotional or practical pain point.
   - Novelty: avoid “4 tips everyone has heard”.
   - Retention path: pattern interrupt every 2-3 seconds.
   - Comment/share trigger.
   - Platform/policy risk.
   - AI-slop risk: visual gibberish, generic stock-feel, impossible UI/text.

4. **Use a human approval loop first**
   - Semi-automatic beats fully automatic at the start.
   - Send drafts/previews to Telegram with: video or script, title, caption, hashtags, duration, target platforms, quality score, and risk notes.
   - When attaching local preview files through Telegram/Hermes, use `MEDIA:/absolute/path/to/file.mp4`; relative `MEDIA:` paths are not reliable for native delivery.
   - Do not post publicly without explicit approval commands such as `Freigabe alle`, `Freigabe YouTube`, `Freigabe TikTok`, `Ablehnen`, or `Ändern: ...`.

5. **Render only after tone is validated**
   - First validate the script voice and niche fit.
   - Then build rendering templates.
   - Otherwise you risk creating polished clips with mediocre text — a beautifully engineered waste of time.

6. **Close the analytics loop**
   - Store per-video metadata and performance: platform, pillar, format, hook, duration, views after 2h/24h/7d, likes, comments, shares, follows, retention if available.
   - Use this to choose the next batch; do not produce blind forever.

## Platform strategy

- **YouTube Shorts:** good primary channel; supports vertical/square Shorts up to 3 minutes. For growth, test 20-45s and selected 60-75s pieces.
- **TikTok:** growth can come from short videos, but Creator Rewards candidates must be 1 minute or longer. Test both short growth clips and 61-75s monetization candidates.
- **Instagram Reels:** strong secondary distribution for the same approved assets; especially good for lifestyle, faith, and save/share formats.
- **Facebook Reels:** useful later for faith/family/life content and older audiences.
- **LinkedIn:** consider only for AI/productivity content, not general FaithTok unless carefully framed.
- **Pinterest:** useful later for evergreen checklist-style content and indirect monetization.

See `references/platform-monetization.md` for concise platform requirements captured from official sources during setup work.

## Visual rules

- Avoid AI-generated images with readable text, fake app UIs, logos, or interface details; these often produce gibberish and reduce trust.
- Prefer real B-roll, screen recordings, kinetic captions, simple cards, and recurring templates.
- First frame must stop the swipe: large hook text, concrete visual tension, or a recognizable human/scene.
- Keep a consistent brand texture: fonts, caption style, color, recurring intro logic, and thumbnail-like first frame.

## Repo setup pattern

For a new shortform automation repo, scaffold around separable concerns:

```text
autoshorts/
  config/
  ideas/
  scripting/
  rendering/
  review/
  publishing/
  analytics/
configs/
  niches/
  platforms/
data/
  drafts/
  performance/
docs/
tests/
```

Commit configs, validation code, draft JSON, and documentation. Do **not** commit `.env`, OAuth tokens, generated credentials, raw downloaded secrets, or generated media outputs.

See `references/autoshortsbot-initialization.md` for a concrete project bootstrap pattern and testbatch schema.

## Pitfalls

- Generic “tips” content is usually too weak unless it has a sharp hook, specific audience, and strong identity.
- 90-second explainer videos often underperform for new accounts; prefer 20-45s growth tests and selected 61-75s monetization tests.
- Do not optimize only for platform payouts early; first find repeatable follower growth.
- TikTok Creator Rewards eligibility can vary by country/account and should be verified in-app before depending on it.
- Do not auto-post during early exploration; use approval-gated previews until the user trusts the channel voice and quality.