Open agent skills for AI video production.
Drop a skill folder into any agent that reads SKILL.md (Claude Code, Cursor, Goose, Codex CLI, …), then describe the video you want to make. The agent walks you through it.
| Skill | What it makes | Validated on |
|---|---|---|
animated-explainer-ad |
9:16 ~30–45s absurdist Pixar 3D explainer ad where a personified villain narrates their own defeat by the product. Bright cartoon characters, single-narrator VO, real-product end card, burned captions, whimsical Pixar/Disney score. ~$15–25 per cut. | HUM Nutrition "Big Chill" · Soteri Skin "Eczema, the pH villain" |
imessage-video-ad |
9:16 ~17–22s iMessage conversation reveal ad — bubbles pop in over time, composer types char-by-char, real Apple iMessage SFX on every send/receive, lofi bed, framed iPhone on a flat-lay desk, brand end card with a promo code. One continuous Playwright recording. ~$0.20–0.65 per cut. | GooseWorks internal builds (product-flex + result-flex) |
More coming. If you ship something using one of these, would love to see it.
Each skill is a self-contained folder:
<skill>/
├── SKILL.md ← agent reads this to drive the workflow
├── README.md ← human-facing intro
├── .env.example ← API keys needed
├── requirements.txt ← Python deps
├── scripts/ ← deterministic glue (FFmpeg + provider SDKs)
├── examples/ ← ready-to-paste kickoff prompts
└── references/ ← worked case studies
The agent reads SKILL.md to know:
- when to invoke this skill (trigger phrases, use cases)
- what inputs to gather from the user
- how to walk the production pipeline phase-by-phase
- when to pause for human review
- what scripts to call and with what arguments
Scripts are kept small, single-purpose, and provider-grounded. No giant config schemas — the agent stitches them.
Paste this into Claude Code (or any agent that can run shell + read files):
Clone https://github.com/gooseworks-ai/goose-video into ./goose-video, then
read the skill at ./goose-video/skills/templates/animated-explainer-ad/SKILL.md
and walk me through it to make an animated explainer ad.
Before starting:
1. cd into ./goose-video/skills/templates/animated-explainer-ad
2. pip install -r requirements.txt
3. Help me create a .env from .env.example — I'll need a FAL_KEY (https://fal.ai)
and ELEVENLABS_API_KEY (https://elevenlabs.io).
Then ask me for: the brand, the product, the problem we're personifying as the
villain, the ownable mechanism, and the audience. From there, follow the SKILL.md
8-phase pipeline (intake → anchors → VO → keyframes → clips → end card → music →
compose) and pause at each human review gate so I can approve before paying for
the next phase.
See each skill's own README for full setup details and additional example prompts.
GooseWorks — AI agents that make videos and run ads. We use these skills internally for client work; publishing them so others can build on the same patterns.
- Shiv Sakhuja — @shivsakhuja · LinkedIn
- Himanshu Bamoria — @0xhbam · LinkedIn
- Akhil Bisht — LinkedIn
- Soham Mehta — @sohamehta_ · LinkedIn
MIT — use freely, attribution appreciated.