Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsAI Video Generator: Shot-by-Shot to Finished Edit
Asking for "a video" usually burns money on long, complex prompts that morph and fail, and leaves you with loose clips.
Free
AI Video Generator: Shot-by-Shot to Finished Edit
Example session with this skill installed
Make a 12-second vertical video ad for our matte black travel mug from this product photo. Text: "Hot for 12 hours" and "Shop now". Upbeat music (I'll provide the track).
- Read your context and instructions
- Compiled the ai video generator
- Generated the image
Shot list (approved)
| # | Dur | Visual | Camera | Mode | Text |
|---|---|---|---|---|---|
| 1 | 3 s | mug on a wooden table, steam rising | slow push-in | Hailuo image-to-video (your photo) | "Hot for 12 hours" |
| 2 | 4 s | mug in a backpack side pocket on a misty trail | tracking, side | Kling text-to-video | - |
| 3 | 3 s | coffee poured into the mug, morning desk light | static | Hailuo image-to-video | - |
| 4 | 2 s | logo + "Shop now" card on brand orange | slow zoom | ffmpeg (no AI cost) | "Shop now" |
Style block: matte black stainless travel mug, warm burnt-orange accents, soft natural light, shallow depth of field, cinematic, calm
Cost
Estimate: 2 Hailuo image-to-video clips (2 x $0.13) + 1 Kling 5 s clip ($0.73) + 1 retake budget ($0.13) =
~$1.12. You approved. Actual:
$1.12 (shot 1 needed the retake).
Takes reviewed
| Shot | Take | Issue | Action |
|---|---|---|---|
| 1 | A | lid shape changed mid-clip | re-run with "keep the mug unchanged, only steam moves" -> B OK |
| 2 | A | strap merged with the mug | kept 1.0-3.8 s segment (clean), trimmed |
| 3 | A | OK | kept |
Edit
python scripts/assemble.py s1.mp4 s2.mp4 s3.mp4 card.mp4 -o mug-ad-9x16.mp4 \
--size 1080x1920 --xfade 0.3 --music track.mp3 --music-vol 0.25 --mute-clips --srt text.srt
saved mug-ad-9x16.mp4 1080x1920 11.4s from 4 clips
- Text kept inside the safe zone (not in the top 12 % or bottom 20 %)
- Loudness -14 LUFS, H.264, faststart (ready for Instagram/TikTok)
Delivered
mug-ad-9x16.mp4 (11.4 s, 8.9 MB), cover.jpg, shot list with exact prompts, cost breakdown, caption + alt text. Want a 1:1 feed version? (re-edit only, no new AI cost)
cover.jpg
PNG · 1536×1024
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
From idea to a finished video, not a pile of random clips
AI video models are expensive per second and unpredictable per take. Asking for "a video" usually burns money on long, complex prompts that morph and fail, and leaves you with loose clips. AI Video Generator plans like a director and edits like an editor.
What it does
- Brief: platform and aspect ratio (9:16, 16:9, 1:1, 4:5), length, subject or product, style, brand, audio, CTA, reference images.
- Shot list first: 3-8 s shots, each with visual, camera move, mode, on-screen text and audio; hook in 2 seconds, CTA at the end.
- Cost control: estimates shots x takes x seconds x price and asks before generating; cheap drafts first, finals only for approved shots.
- No API keys on Agensi: Kling and Hailuo for text-to-video, Hailuo image-to-video for product accuracy from your real photo (from about $0.13 per clip), and Fish Audio text-to-speech for voiceovers. Your agent shows each price and asks first. Also works with Runway, Luma, Veo or Pika via your own account.
- Prompts that work: one subject action and one camera move per shot, style block for consistency, reliable camera moves listed.
- Take review: product correctness, morphing, hands/faces, physics, unwanted text; targeted fixes instead of blind re-rolls; spend tracked against the estimate.
- Finished edit with the included
assemble.py(ffmpeg, tested): normalises size and fps, hard cuts or crossfades, music bed with fades, voiceover, burned-in subtitles, -14 LUFS loudness for social. - ffmpeg recipes (tested): trims, speed ramps, reverse, 16:9 to 9:16 crops, animated CTA cards, seamless loops, cover frames.
- Responsible: no real people or deepfakes, no cloned voices, licensed music only, honest product footage, AI labels where required.
Includes
- SKILL.md workflow
- scripts/assemble.py - clip assembler
- references/shots.md - shot list template, reliable camera moves, prompt examples, pacing by platform, cost tips
- references/edit.md - ffmpeg recipes and caption rules
Pairs with AI Image Generator, Local Text-to-Speech and Remotion Video. Works with Claude Code, Claude (Cowork), Codex, Cursor and any SKILL.md-compatible agent.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Video Skills.
Requires FFmpeg for video assembly. Does not support deepfakes or impersonating real people. Maximum recommended shot length is 8 seconds.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
This skill has been used 7 times.
Trust & safety
Security scan
Security scan passed 2 days ago
- Free with an Agensi account