The /reanimate Skill Teaches an AI Agent to Copy Website Animations with Less Than 1% Pixel Error
The /reanimate skill is a ready-made command for an AI agent that recreates an animation from another website, a screen recording, or a GIF inside your project. The agent records the original as video, writes the code in CSS, GSAP, or Three.js, and compares the frames. According to the author, less than 1% of pixels differ from the original. The agent then fixes any discrepancies on its own. Instead of an evening of manually tweaking delays and curves, you get a working draft in a single run.
AI-processed from TG: @notboring_tech (Not Boring Tech); edited by Hamidun News
You see a beautiful animation on someone else's website: elements glide smoothly into view, a 3D object rotates, the background shimmers. You want to build the same thing, so you open DevTools and fiddle with delays, easing curves, and the order of appearance. An evening disappears, and the result still looks "almost right."
The /reanimate skill (a skill is a ready-made command that you attach to an AI programming agent) takes that grind off your hands: the agent recreates the required animation frame by frame, and the author claims the result differs from the original by less than 1% of pixels. Put simply, if you overlay the original frame and the copy, fewer than one hundredth of the image's points will differ.
How It Works
The idea is similar to the work of a careful front-end developer who has an assistant with a good eye. You give the agent a source: a website link, a screen recording, or a GIF. The agent records a video of the target animation, writes code that reproduces it in your project, and runs it.
Next comes the most useful part: the agent records a video of its own version and compares it with the original frame by frame. If the picture diverges somewhere, for example an element appears a few milliseconds late or a gradient's color has shifted, the agent finds the error and edits the code. The cycle of "write, record, compare, fix" repeats without human involvement.
This approach makes sense because the agent gets an objective check. Usually an AI writes animation code "by eye" from a text description and has no way of knowing whether it looks right. Here there is a reference to compare against, pixel by pixel.
Which Technologies It Supports
According to the author's description, the skill generates code using several popular tools:
- CSS for simple transitions and keyframes;
- Framer Motion for React projects;
- GSAP for complex timelines;
- Three.js for 3D scenes.
The author also states that the skill handles harder cases: complex 3D animations, Canvas, WebGL, and WebGPU. These effects are the hardest to reproduce by hand, because they have no tidy CSS rules you can read in an inspector. Everything happens inside shaders and render loops. That is why the approach of "look at the result and adjust the code to match" is especially useful here.
The "less than 1% of pixels" figure comes from the author's post and the repository page. The material contains no independent measurements. It is unclear what set of animations was used for the test, and how well the result holds up for long or interactive effects that depend on scrolling and cursor movement. Before building a workflow around the skill, run it on two or three of your own examples.
Why It Matters for Developers and Product Managers
For practitioners, there are three scenarios.
The first is rapid prototyping. A designer brings a reference in the form of a screen recording, and instead of a long discussion about "how to build this," the team gets working code from a single agent run.
The second is porting animations between technologies. If an effect was built on one stack and the project runs on another, the skill can rebuild it based on the video rather than the original source code.
The third is a lesson for anyone building agents for visual tasks. The principle of "record, compare against the reference, fix" works beyond animation: give an agent a way to measure how close its result is to the target, and quality improves.
There is also a question of rights. Technically reproducing an animation does not mean you have the right to copy it. If you take a reference from someone else's website, make sure this does not infringe on others' rights or turn into a clone of a distinctive brand style. The tool only speeds up the work; responsibility for how it is used stays with the team.
The project repository is open on GitHub, so you can review and verify everything yourself. For front-end teams, this is a good reason to try the agent skill on a real task and measure how much time it saves compared with manual tweaking.
- Nov 21, 2026Apple Is Building an AI Agent and Making It Harder for Outsiders to Get Access: Permission Now Takes 14 Steps, Not 4
- Nov 21, 2026The Programmer Stopped Being an AI Agent Dispatcher: Multica Takes Over the Handoff Routine
- Nov 21, 2026Developer Skips Pinecone: Searching 72,450 Code Chunks in 50 ms
- Nov 21, 2026Eight Open-Source Tools for AI Agents: Browser Use Leads with 121,000 Stars
Key Questions Answered
How It Works?
The idea is similar to the work of a careful front-end developer who has an assistant with a good eye. You give the agent a source: a website link, a screen recording, or a GIF. The agent records a video of the target animation, writes code that reproduces it in your project, and runs it. Next comes the most useful part: the agent records a video of its own version and compares it with the original frame by frame.
Which Technologies It Supports?
According to the author's description, the skill generates code using several popular tools: CSS for simple transitions and keyframes; Framer Motion for React projects; GSAP for complex timelines; Three.js for 3D scenes. The author also states that the skill handles harder cases: complex 3D animations, Canvas, WebGL, and WebGPU. These effects are the hardest to reproduce by hand, because they have no tidy CSS rules you can read in an inspector. Everything happens inside shaders and render loops.
Why It Matters for Developers and Product Managers?
For practitioners, there are three scenarios. The first is rapid prototyping. A designer brings a reference in the form of a screen recording, and instead of a long discussion about "how to build this," the team gets working code from a single agent run. The second is porting animations between technologies. If an effect was built on one stack and the project runs on another, the skill can rebuild it based on the video rather than the original source code.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.