Agent skills are machine-readable SKILL.md files that tell AI coding agents how to execute specific production tasks. As of March 2026, the ecosystem has expanded well past code generation into specialized video workflows. This guide maps 18 skills to the 9 stages of an AI video pipeline, with install commands and compatibility notes.
If you already use Claude Code, Codex CLI, or Cursor, these agents can handle scripting, image generation, video editing, audio production, and publishing. Each skill below is a standalone file you install in under a minute.
Related: Read the complete AI video pipeline guide, browse the skills directory, or see how agent skills automate video production.
TL;DR: Quick Reference Table
| Stage | Recommended Skill | What It Does | Install |
|---|---|---|---|
| Script | aivp-script | Structured scripts with scene breakdowns | npx skills add aivp-script |
| Storyboard | aivp-storyboard | Scripts to visual shot lists | npx skills add aivp-storyboard |
| Image | fal-generate | Image generation via fal.ai (FLUX, SD3) | npx skills add fal-generate |
| Video | ai-video-generation | Multi-model video clips via inference.sh | npx skills add ai-video-generation |
| Edit | ffmpeg-editing | FFmpeg trim, concat, overlay, encode | npx skills add ffmpeg-editing |
| Audio | elevenlabs-voice | Voiceover and voice cloning via ElevenLabs | npx skills add elevenlabs-voice |
| Metadata | seo-optimizer | SEO titles, descriptions, schema markup | npx skills add seo-optimizer |
| Publish | social-content | Platform-specific clips for Shorts/Reels/TikTok | npx skills add social-content |
| Review | aivp-review | Automated quality checks | npx skills add aivp-review |
What Are Agent Skills?
An agent skill is a SKILL.md file: a Markdown document with frontmatter and plain-language instructions that an AI coding agent parses and executes. The format was introduced by Anthropic for Claude Code in early 2026, then adopted by Codex CLI, Cursor, and Gemini CLI. Write once, run on any compatible agent.
Skills differ from shell scripts because the agent interprets instructions contextually: adapting to missing inputs, chaining API calls, and making decisions within defined boundaries. Installing is one command: npx skills add [skill-name]. You can also install from GitHub, skills.sh, or ClawHub. See agentskills.io for a deeper walkthrough.
1. Script
Transforms a topic into a structured script with narration, scene descriptions, and timing.
aivp-script generates pipeline-format scripts: structured JSON with scene breakdowns, narration, visual directions, and duration targets.
npx skills add aivp-scriptcreate-viral-content focuses on hook-driven short-form scripts with attention-grabbing openings and platform-tuned calls to action. Available on skills.sh.
npx skills add create-viral-content2. Storyboard
Converts a finished script into a visual shot list. Investing here reduces expensive re-generations at the Video stage.
aivp-storyboard takes a structured script and produces a shot-by-shot storyboard with visual descriptions, camera movement, and estimated duration.
npx skills add aivp-storyboardcanvas-design generates branded visual assets: thumbnails, social cards, and storyboard frames with brand-consistent styling.
npx skills add canvas-design3. Image Generation
Produces visual frames that become video keyframes or standalone assets.
fal-generate connects to fal.ai's inference API for FLUX.2, SD3, and other models through a single interface.
npx skills add fal-generateaivp-image is the pipeline-native image step: reads storyboard output, generates images per scene, saves in the correct directory structure.
npx skills add aivp-image4. Video Generation
Converts images or text prompts into video clips. Typically the most expensive stage in API credits.
ai-video-generation from inference.sh provides multi-model generation through a unified API: Seedance, Kling, Minimax, and others.
npx skills add ai-video-generationaivp-video reads previous stage outputs, generates clips per scene, and organizes them for editing. Supports text-to-video and image-to-video.
npx skills add aivp-video5. Editing
Assembles clips, applies transitions, overlays text, and exports the final cut.
ffmpeg-editing exposes FFmpeg operations through natural language: trim, concatenate, overlay, adjust speed, add watermarks, encode.
npx skills add ffmpeg-editingvideo-toolkit from DigitalSamba provides higher-level editing: multi-track assembly, audio/video sync, subtitle burn-in, platform export presets.
npx skills add video-toolkit6. Audio
Handles voiceover generation, music selection, and mixing.
elevenlabs-voice generates voiceover via the ElevenLabs API with dozens of voices, voice cloning, multilingual output, and synced SRT subtitles.
npx skills add elevenlabs-voiceaivp-audio reads the script, generates voiceover per scene, selects background music, mixes tracks, and outputs Edit-ready files.
npx skills add aivp-audio7. Metadata and SEO
Generates titles, descriptions, tags, and structured data for discoverability.
seo-optimizer produces SEO-optimized video page metadata: titles, descriptions, Open Graph tags, Schema.org markup, and keyword strategies.
npx skills add seo-optimizertranscript-fixer cleans auto-generated transcripts, fixes punctuation, and structures text for descriptions, blog posts, or closed captions.
npx skills add transcript-fixer8. Publishing
Handles platform-specific formatting, uploading, and cross-posting.
social-content reformats finished video into platform-optimized versions with cropping, captions, hashtags, and upload packages for YouTube, TikTok, Instagram, and X.
npx skills add social-contentyoutube-clipper extracts highlight clips from long-form videos, generates Shorts-optimized vertical crops, and creates metadata packages.
npx skills add youtube-clipper9. Review
Runs quality checks before publication: resolution, audio clipping, metadata, format compatibility.
aivp-review validates resolution, frame rate, audio levels, file size, and metadata completeness. Produces a pass/fail report.
npx skills add aivp-reviewprompt-architect reviews and optimizes prompts used across the pipeline for clarity, specificity, and model compatibility.
npx skills add prompt-architectHow to Install and Manage Skills
# 1. Install your agent
npm install -g @anthropic-ai/claude-code
# 2. Create skills directory
mkdir -p .claude/skills
# 3. Install skills
npx skills add aivp-pipeline
npx skills add ffmpeg-editing
npx skills add elevenlabs-voice
# 4. Invoke a skill
claude
> /aivp-pipeline topic="Product demo video" platform="youtube" style="professional"For full 9-stage orchestration, use the aivp-pipeline master skill.
Compatibility Matrix
| Skill | Claude Code | Codex CLI | Cursor | Gemini CLI |
|---|---|---|---|---|
| aivp-script | Yes | Yes | Yes | Yes |
| aivp-storyboard | Yes | Yes | Yes | Yes |
| fal-generate | Yes | Yes | Yes | Partial |
| ai-video-generation | Yes | Yes | Partial | Partial |
| ffmpeg-editing | Yes | Yes | Yes | Yes |
| video-toolkit | Yes | Yes | Yes | Partial |
| elevenlabs-voice | Yes | Yes | Yes | Yes |
| seo-optimizer | Yes | Yes | Yes | Yes |
| social-content | Yes | Yes | Partial | Partial |
| youtube-clipper | Yes | Yes | Yes | Partial |
| aivp-review | Yes | Yes | Yes | Yes |
| prompt-architect | Yes | Yes | Yes | Yes |
"Partial" means the skill runs but may lack full tool access on that platform.
FAQ
What is an AI agent skill for video production?
A SKILL.md file with machine-readable instructions for a specific video production task. An AI coding agent reads it and executes autonomously: calling APIs, writing files, running commands, and chaining outputs between stages.
How do I install agent skills?
Run npx skills add [skill-name]. This downloads the SKILL.md into .claude/skills/. Also available from GitHub and skills.sh. Invoke with a slash command like /ffmpeg-editing.
Do skills work with Cursor and Codex, or only Claude Code?
The SKILL.md format is agent-agnostic. Claude Code, Codex CLI, and Cursor all support it. Gemini CLI has partial support. Most skills without browser-action requirements work across all four.
Can I chain multiple skills into a pipeline?
Yes. aivp-pipeline chains all 9 stages, with each output feeding the next. You can also create custom orchestration skills referencing sub-skills in a master SKILL.md.
Are agent skills free?
Skills are free and open source. External APIs they call (ElevenLabs, fal.ai, inference.sh) have their own pricing. A 5-second clip via inference.sh costs $0.05-0.30 depending on the model.
How do I create a custom skill?
Write a SKILL.md with YAML frontmatter (name, description) and a Markdown body with execution steps. Place in .claude/skills/your-skill-name/SKILL.md. See the agent skills automation guide for examples.
