A weak prompt ("a woman drinking coffee") gives you a weak video. A good prompt has six things: subject, action, scene, camera, style, and what to avoid. You don't have to learn to write those. The prompt enhancer skill builds all six into every prompt for you, so you describe the vibe and Claude writes the full cinematic version.
Download the skill, then upload it into Claude:
⬇️ Download the prompt enhancer skill
Prefer to build it yourself? Here's the whole skill, paste it into a SKILL.md file:
---
name: video-prompt-enhancer
description: Turn a rough one-line video idea into a full cinematic prompt (subject, action, scene, camera, style, and what to avoid), ready to generate. Use whenever the user wants an AI video and gives a vague or short description, asks to "make a video", "enhance this prompt", or wants to generate through Arcads.
---
# Video Prompt Enhancer
When the user gives you a rough video idea, expand it into ONE detailed, photoreal cinematic prompt built from these six elements. Never hand back the user's thin one-liner: always return the enhanced version.
1. Subject: who or what is on screen (age, look, wardrobe, defining details).
2. Action: what they are doing, moment to moment.
3. Scene: the setting, time of day, background, atmosphere.
4. Camera: shot type and movement (slow dolly-in, handheld close-up, macro, orbit).
5. Style: the look (cinematic, film grain, colour palette, lighting, mood, lens).
6. What to avoid: negatives (no on-screen text, no warped hands, no extra limbs, no logos).
Ask ONE clarifying question only if the idea is too vague to place a subject or scene. Otherwise make tasteful, concrete choices and proceed. Output one clean, photoreal paragraph with all six elements woven in. Avoid empty adjectives like "beautiful". If the Arcads connector is enabled and the user asks to generate, send the enhanced prompt to Arcads (it runs Veo 3.1, Sora, Kling; turning on Fable 5 improves realism).You need something to actually render the video. Connect Arcads once:
Now just ask Claude to use the skill and Arcads together:
Use the video prompt enhancer skill to turn this into a full cinematic prompt, then send it to Arcads to render: [your rough idea, e.g. "a cosy morning coffee ad, warm light, slow motion"].Claude writes the full cinematic prompt and sends it straight to Arcads. It gets even better with Fable 5 turned on. Arcads runs the best models in one place, so it can call Veo 3.1, Sora, and Kling for you.
🤖 Claude: claude.com (add the skill under Customize, then Skills)
🎬 Arcads: arcads.ai (add it under Customize, then Connectors)