Cindy Zhu.
Free guides
Content Claude

Turn Claude into your entire content team

Hey, it's Cindy ๐ŸŒฑ You commented TEAM, so here is the exact setup: a skill that makes your graphics, a video framework that renders your animations as code, and a connector that generates your B-roll.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
the shape of it

What you are actually wiring up

Three roles you would otherwise hire, plus the plumbing that connects them. You do not need all of it on day one, and each piece is useful on its own.

Design The canvas-design skill. You describe a graphic, Claude produces real PNG and PDF files rather than a description of a graphic.
Video Remotion. Claude writes React code that renders to MP4, so your intro animation becomes a reusable file you tweak instead of rebuild.
Visuals Higgsfield, connected to Claude. Claude writes the shot prompts and generates the clips you would otherwise have filmed.
The wiring MCP, the open standard that lets Claude talk to outside tools directly instead of you copying output between tabs.
Honest framing before you build: this replaces the production work, not the taste. Claude will make the graphic, render the animation and generate the clip. Deciding what is worth making, and whether the result is good, stays yours. That is the part that makes your content yours.

piece one

Design: install the canvas-design skill

A skill is a folder with a SKILL.md inside that teaches Claude a process once. canvas-design is Anthropic's own, and it is specifically built to output finished .png and .pdf artwork instead of a text description of a design. Uploading a skill works on every plan, Free through Enterprise.

  1. ๐ŸŒฑ Download the skill. Open github.com/anthropics/skills, click the green Code button, then Download ZIP. Unzip it and find the skills/canvas-design folder inside.
  2. ๐Ÿ’ก Zip that one folder on its own. Right click canvas-design and compress it, so you end up with a single canvas-design.zip containing the folder and its SKILL.md. Uploading the whole repo zip will not work.
  3. ๐Ÿ‘ฉโ€๐Ÿ’ป Upload it to Claude. In the Claude app or on claude.ai, open Customize, then Skills, then Add, then Upload a skill, and pick your zip. If you use Claude Code instead, skip the zipping and drop the canvas-design folder straight into ~/.claude/skills/.
  4. ๐Ÿง  Brief it like a designer. Ask for the design direction before the file, approve it, then let it export. Skipping the direction step is how you end up regenerating the same graphic nine times.
๐Ÿ’ก brief the designer
Use the canvas-design skill. I need three Instagram carousel covers for [topic], sized 1080 by 1350, in my brand: a cream background, one warm orange accent, a serif headline with a clean sans body, generous white space, and no stock photography.

Each cover needs a short headline of five words or fewer, a small lowercase kicker label above it, and calm empty space at the bottom where my handle will sit. The three should read as a set, so keep the type scale and the accent placement consistent.

Give me the design direction first, in one short paragraph, so I can approve it before you make anything. Once I say go, export each cover as a full resolution PNG and give me a single PDF with all three.
Only install skills from sources you trust, since a skill is a set of instructions Claude will follow. Anthropic's own repo is about as safe as it gets, which is why it is the one to start with.

piece two

Video: your intro animation as code

Remotion lets you build videos in React, which sounds like a developer thing until you realise what it means for you: your intro animation stops being a file you re-export from an editor and becomes something Claude can edit on request. Change the title, change the colour, re-render, done.

Plan gate, stated plainly. Remotion's Free License covers individuals and companies of up to three people, with unlimited commercial use. Four or more people needs a paid Company License. Terms and current pricing are on Remotion's pricing page. You also need a coding agent, and most of those are paid.

  1. ๐ŸŒฑ Install Node.js. Grab the LTS version from nodejs.org. This is the one prerequisite, and it is a normal installer with a next button.
  2. ๐Ÿ‘ฉโ€๐Ÿ’ป Create a blank project. In your terminal, run npx create-video --yes --blank my-video. That builds a fresh Remotion project in a folder called my-video.
  3. ๐Ÿ’ก Set it up. Run cd my-video, then npm install, then npx remotion skills add. That last command installs Remotion's own agent skills so Claude knows the framework properly instead of guessing at the API.
  4. ๐Ÿง  Start the studio. Run npm run dev. A live preview opens in your browser and updates every time the code changes, so you watch the animation build itself.
  5. ๐ŸŒฑ Point Claude at the folder. Open a second terminal window, cd my-video, and start Claude Code with claude. It can now read and edit the project files directly.
  6. ๐Ÿ’ก Render the MP4. When you like it, ask Claude for the render command, or run npx remotion render yourself and pick the composition. The MP4 lands in the out folder.
๐Ÿ‘ฉโ€๐Ÿ’ป build my intro animation
Build me a reusable intro animation in this Remotion project. Three seconds, 1080 by 1920, cream background, my handle appearing letter by letter in a serif font, a thin orange line drawing itself underneath as the letters land, and a soft fade out over the final 15 frames.

Make the handle text, the two colours and the total duration into props on the composition, so I can change them later without touching the animation code. Then add a second composition that renders the same animation at 1920 by 1080 for landscape.

When it works in the preview, tell me the exact command to render each one to MP4, and explain in two lines which file I edit if I want to change the timing later.
The reason this beats an editor template: next month you say "same intro, new title, make the line thicker" and it takes one sentence. Templates make you redo the export. Code makes you change one prop.

piece three

Visuals: connect Higgsfield to Claude

This is the B-roll department. Once Higgsfield is connected, Claude writes the shot prompts and generates the clips inside the same conversation, so you brief a scene instead of tab-hopping. You connect it once.

  1. ๐ŸŒฑ Copy the connector link. Go to higgsfield.ai/mcp and copy the connector link. The link you paste is mcp.higgsfield.ai/mcp.
  2. ๐Ÿ’ก Add a custom connector. In Claude, open Customize, then Connectors, then Add custom connector.
  3. ๐Ÿ‘ฉโ€๐Ÿ’ป Name it and paste. Call it Higgsfield, paste the link, then click Add and Connect.
  4. ๐Ÿง  Authorise it. Sign in with your Higgsfield account. This one is paid, and video generation needs a plan with credits, so set that up before you connect or the first request will fail.
  5. ๐ŸŒฑ Optional power-up. Upload Higgsfield's own skill file so Claude knows how to drive the engine properly. It lives at github.com/higgsfield-ai/skills, and it installs the same way as the design skill above.
๐ŸŒฑ generate the B-roll
Using the Higgsfield connector, generate four B-roll clips for a video about [topic]. Each one is five seconds, vertical 9 by 16, and should read like real footage rather than an obvious render.

Here are the shots: [shot 1, with the subject, the lighting and the camera move], [shot 2], [shot 3], [shot 4]. Keep the look consistent across all four, same colour temperature, same lens feel and same depth of field, so they cut together in a single sequence.

Show me the four prompts you are about to send before you spend any credits, so I can adjust them. Then generate them one at a time and give me the download link for each.

For still images rather than motion, Reve is the image model I use when a cover or thumbnail has to carry the post. Also paid, and worth it only once the graphic is doing real work.


the wiring

MCP is what makes it one team

MCP stands for Model Context Protocol. It is an open standard for connecting Claude to outside tools, and it is the difference between three separate AI tools and one team. Without it you are the integration: you copy a prompt out of Claude, paste it into another tool, download the result, and bring it back. With it, Claude calls the tool itself and keeps the context. Setup and troubleshooting live in Anthropic's custom connectors guide.

In the Claude app Skills and connectors both work here, so design and Higgsfield visuals run in one chat. Your Remotion project lives on your computer, so the app cannot touch it. Best if you want two thirds of the team with zero terminal.
In Claude Code Skills, connectors and your local project files all sit in the same place, so the whole team runs from one brief. This is the version where you say "make this week's content" and it makes graphics, updates the animation and generates the clips. Best if you are willing to open a terminal.
๐Ÿง  one brief, three tools
Here is this week's content brief: [topic, the hook, and the three or four key points]. Run it end to end using the tools you have.

First, use the canvas-design skill to make the carousel covers and the in-post graphics. Then, in my Remotion project, update the intro animation with this week's title and render it. Then use the Higgsfield connector to generate the three B-roll clips the script calls for.

Before you start anything, list exactly what you are about to make, ask me for anything you are missing, and flag every step that will cost credits so I can approve it first. Work through them one at a time and show me each result before moving on.

where to start

Do not build all three this week

Pick the job you currently hate most. If you are burning hours in Canva, install the design skill and stop there. If your intro is the bottleneck, do Remotion. If you are hunting for stock footage that never fits, connect Higgsfield. Add the next piece only when the first one has genuinely saved you an hour, because a half-built system you do not trust is slower than doing it by hand.

A realistic first week: install canvas-design, brief it on one real carousel, and see whether the output survives your own taste test. That single loop tells you more about whether this setup fits you than reading another ten guides. ๐ŸŒฑ

Sources

canvas-design skill, Anthropic's skills repo
Agent Skills, official overview
Remotion, prompting videos with coding agents
Remotion licensing and pricing
Custom connectors with remote MCP (Claude Help Center)

the waitlist is now open

AI System Club is launching soon

Build AI systems with me through recorded lessons you can follow at your own pace. Join the waitlist for first access and the founding price.

btw this list gets first dibs
the moment doors open :)