📰 Blog guide

4 AI Agent Skills That Actually Auto-Produce Short-Form Videos (2026 Tested)

CreatorStack Team July 8, 2026

The Problem With AI Video Tools

Short-form creators face a paradox: AI tools are supposed to save time, but most workflows still need 5–6 different apps. ChatGPT for script, CapCut for editing, ElevenLabs for voice, a third-party subtitle tool, a stock footage site — switching between them kills momentum.

Agent Skills solve this. A Skill is a markdown instruction file that tells an AI agent (Claude Code, OpenClaw, Cursor, Windsurf, Codex CLI) exactly which tools to call, in what order, and how to handle errors. You describe what you want in natural language; the agent runs the full pipeline.

I spent two weeks testing the 4 most practical open-source Skills for short-form video creators. Here’s what I found.

The 4 Skills That Make the Cut

Author: Aaron He Zhu
Works with: Claude Code, OpenClaw, Cursor, Windsurf
GitHub: aaron-he-zhu/seo-content-writer

This Skill automates the research + outline + first draft pipeline. You tell it your topic and target platform, and it runs a structured workflow:

Identifies focus keyword and long-tail variants
Analyzes top-10 SERP results for content gaps
Builds an H1–H2–H3 outline with proper hierarchy
Ensures core answer appears in the first 150 words
Adds at least one external citation per 500 words
Outputs a complete draft with TL;DR summary

Real test: I asked it to write a script for “Best AI Tools for YouTube Thumbnails 2026.” It returned a 1,200-word script with 3 hook variants and a CTA section. Edits needed: about 15 minutes to add personal opinions.

Best for: Long-form scripts, newsletter deep-dives, platform-specific content (Xiaohongshu, WeChat Official Accounts).

Limitation: The output is text; you still need a separate tool to turn it into video.

2. BibiGPT Skill — Instantly Summarize Any Video

Author: JimmyLv (BibiGPT founder)
Works with: Any AI agent (universal Skill)
GitHub: JimmyLv/bibigpt-skill

This is the Skill I used most during testing. It extracts captions from over 30 video platforms (Bilibili, Douyin, YouTube, Xiaohongshu, Xiaoyuzhou podcasts, local MP3/MP4 files) and returns a structured summary with timestamps.

Key features:

Platform coverage: 30+ platforms including Chinese ones (Bilibili, Douyin, Xiaohongshu)
Caption sources: Bullet comments (danmaku), CC subtitles, auto-generated captions
Output flexibility: --lang en for English summaries, --summary-only for brevity, --chapters for section breakdown
Response time: ~30 seconds for a 30-minute video

Real test: I fed it a 25-minute YouTube tutorial on AI video editing. It returned a 300-word summary with 7 timestamped chapters, each with a one-sentence description. I then used that summary to decide which 3 minutes of the video to watch — saved me 22 minutes.

Best for: Researching competitor content, repurposing long-form into short clips, staying current on industry trends without watching full videos.

Limitation: It summarizes existing content; it doesn’t create new videos.

3. Narrator AI CLI Skill — Full Video Production From One Sentence

Author: AI解说大师 (AI Narration Master)
Works with: OpenClaw (requires companion CLI)
GitHub: jieshuo-ai/narrator-ai-cli-skill (+ companion CLI)

This is the most ambitious Skill in the list. It takes a single sentence like “Make a movie commentary video for Pegasus in comedic style” and runs an 8-step production pipeline:

Searches for movie resources
Selects a commentary template
Matches background music
Picks a voice actor (63 roles available)
Generates narration script
Syncs audio to visuals
Renders final video
Returns a download link

Resource library: 93 movie files, 146 BGM tracks, 63 voice roles, 90+ commentary style templates.

Setup:

# Install companion CLI
pip install narrator-ai-cli

# Copy SKILL.md to agent skills directory
cp narrator-ai-cli-skill/SKILL.md ~/.openclaw/skills/

Best for: Creators already running movie/short-drama commentary channels who want to scale from 2 videos/day to 20.

Best for (beginners): Anyone who wants to start a commentary channel but is blocked by video editing skills.

Limitation: Requires an API key (the author provides access in the README). The companion CLI is mandatory — the Skill is the “brain” but the CLI is the “hands.”

4. Remotion Best Practices Skill — Code-Generated Videos

Author: Remotion team (open-source)
Works with: Any agent (Universal Skill)
GitHub: remotion-dev/skills

This Skill deserves a spot because it solves a problem no other tool can: deterministic video generation from code.

Remotion is an open-source framework that renders React components to MP4. The official Skill includes 30+ rule documents covering animation, audio, subtitles, 3D, charts, transitions, Lottie animations, font loading, video frame extraction, and TikTok-style subtitle effects.

Once installed, the agent can generate fully renderable video code from a prompt:

Use Remotion best practices to create a 30-second 1080×1920 educational video
about how AI agents work, split into 5 scenes. 
- Text safe zone: top 150px, bottom 170px
- Main title: 56px minimum
- Style: clean, dark background, accent blue

The agent will:

Research the topic
Write a 5-scene script
Generate React animation code per scene
Open Remotion Studio for preview

The magic: Because the output is code, you can iterate deterministically. “Widen the bar chart” → “Increase font size” → “Switch to orange palette” — each change is predictable, not a random regeneration like AI video tools.

Best for: Knowledge content (educational), data visualization, template-based batch production.

Limitation: Requires some technical comfort (code, React, CLI). Learning curve is steeper than the other three Skills.

The One Workflow I’d Ship Today

After testing all four, here’s the stack I’d recommend for a creator automating short-form video:

Research competitors   → BibiGPT Skill  (5 min)
Write script           → SEO Content Writer Skill  (10 min)
Generate video         → Narrator AI CLI Skill  (3 min)
Batch high-end content → Remotion Best Practices Skill  (15 min/template)

Total time per video: under 20 minutes (versus 3-6 hours manually).

What the creator still does:

Choose the topic and angle
Review and inject personal opinions
Hit publish and engage with comments

How to Install These Skills

Different agents use slightly different paths:

Claude Code:

# Copy SKILL.md to skills directory
npx skills add <github-repo>
# Or manually: cp to ~/.claude/skills/

OpenClaw:

# Search and install from ClawHub.ai
# Or CLI:
npx clawhub@latest install <skill-name>
# Or manual: cp to ~/.openclaw/skills/

Cursor / Windsurf / Cline:

npx skills add <github-repo>

Codex CLI: Use the built-in skill-installer tool.

Pro tip: After installing, restart your agent. Then test with a minimal prompt first (e.g., “summarize this video URL”) before running full production.

Which One Should You Install?

If you…	Install this
Write scripts for platforms	SEO Content Writer
Need to research competitor videos	BibiGPT Skill
Run a movie/short-drama channel	Narrator AI CLI Skill
Want code-level control over video quality	Remotion Best Practices Skill
Want all of the above	Install all 4 — they complement each other

The Bottom Line

Agent Skills are not a gimmick. These 4 tools represent a genuine leap in creator productivity — not because AI replaces creativity, but because it removes the mechanical overhead between having an idea and publishing a video.

The creator who installs these Skills today is running a production pipeline that would have required a 3-person team in 2024.

Disclosure: Some links above are affiliate links. I test every tool independently and only recommend what actually works.