Claude Cowork + Skills is Your NEW Content System

If you a marketer, you know how much of the week goes into just creating content, but today we are going to change that.

I'll show you how to build your content system using Claude work and Claude code that handles repurposing carousel post as, as your content, and even your content calendar.

All from one single project folder, multiple platforms, but one system.

Let's go.

Every content system should have at least three core components.

The context layer, this is your brand knowledge, your voice, your audience, your products.

This is what makes the output sounds like you.

The skill layer, the reusable instructions that tells the system how to execute specific task.

The orchestration layer Claude with your tools connected, reading your inputs, running the right skills, and pushing content outputs.

And the whole system can be controlled with something as simple as just a spreadsheet.

For this demo, the brand is called Goldie Coffee, which is a specialty cafe brand, and inside we have three parts, what I call the system folder that store the useful brand context, the skills just for this project and a script folder.

And then your working folder for your marketing content like Ads, SEO, social, and most importantly, the Cloud MD file.

That is the system prompt for this project and the content planner Excel, which I'll explain in a minute.

And by the way, this is what we are covering inside our AI growth System for marketing program.

You will learn the complete setup and get all the templates to design your own.

So if you're interested, you can find a link below to join.

Now before we even get started, Claude Cowork doesn't automatically read the project level skills like Claude Code does.

So you will want to add these instructions to your Claude.MD file to point it to your skill folder, or if you want this to apply across all your projects, you can just set it once on the settings, cowork global instructions, using the same Prompt instruction.

So either method will work.

Okay.

Our first use case is the one every marketer deal with content repurposing, and there is a reason we start here.

This is the highest ROI task because you are not creating anything new.

You are just unlocking more value from content you have already made.

And using just one system, you can turn one piece of content into multiple formats.

So here inside the social input folder.

So we already have three blog posts that we want to repurpose.

So think of this where you put all the files as part of the input for AI to process and to create or repurpose content to multiple formats as the outputs.

Now on Claude cowork pick the same project folder.

So we will use this prompt to ask it, to read the blog content and then repurposing them into different social content since we will also generate visuals.

So make sure you also install the nano banana MCP so it is now loading the relevant context, including the Claude.MD file, as well as the project specific skills and the post creation framework.

And you'll notice it spinning up parallel agents automatically.

So that is cowork doing it for you.

So 99% of time you don't need to create subagent manually inside Claude cowork.

And pretty quickly it comes back with a total 9 posts and six visuals all repurpose from the same set of sources.

So each post indicate the source use, the format and even the hook and also the visual direction in case you need to pass to a teammate to handle the visual.

And also the LinkedIn posts with the post structure and hook extracted from the framework and the newsletter draft with the subject line options and suggested CTA.

Of course, we still need to review it to make sure it aligns with our voice before publishing.

Now for the social visuals, they are also pretty decent and they're generated to align with the post content, and I have intentionally asked it to exclude the text.

And all of these files are saved at inside the output folder.

And so this is the power of setting up your content system right from the beginning to keep everything organized.

So we just saw how this content system can do and everything we set up lives right inside Claude cowork.

And if you want to take this further, I recommend this free resource from HubSpot, the Claude Cowork Stack 12 advanced Prompt to replace a week of work.

It's a great compliment to what I cover in this video.

This is the resource that gives you 12 structure prompts built specifically for how Cowork works with your files and connectors, which is very helpful for your content workflows like the content performance analyzer, competitive intelligence report, content repurposing pipeline.

What I particularly like is how these prompts are all designed around the most common workflows every marketers would need, and a great inspiration to design my own Claude skills and prompts.

So start with the templates and then adapt them for your specific workflows so you can build your own marketing prompts and skill library over time.

You can grab it for free below, and thank HubSpot for sponsoring this video.

All right.

We just built a repurposing content system, but what about creating new content, especially the one that drives the most engagement Carousel post, and this is one of the most popular content format every marketer loves.

Using Claude, we can build a carousel post generation system that generates full on brand carousel sets, visuals and copies right inside Claude Cowork based on your input brief.

And this is just a basic template and this system can generate much more than that.

So to generate beautiful on brand carousel visuals, the key is to set up what I call a style library.

So inside this example folder, we have this carousel style library where it contains different carousel post style.

So you include different distinct style to give AI inspiration direction.

But very important, we are not asking AI to follow this template rigidly, but to use it as a guidance only.

And also this style guide file lets Claude to quickly grasp what is inside so it doesn't have to analyze every image in real time.

Then we also have this visual folder that will handle all the visual input and outputs, like the example I just showed you.

And here we have the content planner.

So think of it as your command center.

So each tab defines the input requirement for that content type.

And the first one is already done from our previous demo.

So everything is ready now on cowork pick the same project folder.

So use this prompt to ask it to read through the content planner and then to create a carousel post for those that have the Emplty status and to update the content planner after complete.

So immediately it read, there are five sets of carousel posts need to create and launch five carousel generation tasks in parallel and calling the carousel post designer skill.

And pretty quickly we have all 23 carousel visuals being generated for different platforms. So they are grouped by a campaign topic with the generator post copy.

And this is quite impressive, is not exactly following the template exactly what we want, but adding its own variation to make it look fresh and on brand.

So some are more text heavy style, while some are visual heavy with different vibe.

But I see the logos are not consistent as we haven't given it our brand logo.

So this is the area we can enhance.

But you can see these carousel post draft are pretty good, even though you just want to use it as a quick mockup or a client brief.

And the content planner has also been updated with the output file paths, the complete status.

So this is why even with just a spreadsheets, everything is working like a system.

Now we can take a step further to create a scheduled task right inside the same conversation.

So we can use the built-in slash command schedule to set up a new scheduled task to let it run this whole workflow every Friday at 9:00 AM then you can review the details and proceed if you like it.

And you can always make changes to the task by checking all the existing created schedule task.

But an important note is that for now, the scheduled task function will only run while your Claude desktop is open.

And my tip is to create a schedule task that doesn't require human interactions or approval.

So the agent can really run itself.

Now at this point, we have got carousel Post generating themself on a schedule, but all of that content is still sitting in your local folders.

Your team can't see it, can't review it, can't publish it.

And this is why your system needs a distribution layer, a way to automatically push finished content where people can actually review, approve, and publish it.

So let me show you how that works.

So for this task, since we also need to upload media files, like the social visuals to our Notion database.

So we will create a reusable Python script inside the system script folder.

Just think of this script as a mini program.

So Claude will use whenever it needs to upload the media files to Notion using the API key.

So going back to our content calendar, we are going to upload all the files from our output folder to Notion and then to ask Claude to update the notion schedule status once finished.

And since we need to upload media files through external API calls, so for that purpose we will need to use Claude code.

So now go back to Claude desktop.

From there we'll switch to Claude code and again here, pick the right project folder and then make sure we already got the notion MCP connected.

Then use this prompt to ask it, to read the content planner and to create a new entry on the notion content calendar for those where the schedule on notion status is empty.

So Claude would spin up agents to run this task and set up the to-do list for the notion entry creation and social visual uploading.

So in just one second, Claude, push all the entries to notion where you can see side by side.

And finally updating the content plan with the spreadsheet.

So now all the six new carousel posts are already being created on our calendar.

So all posts are created correctly.

So you can see there is a title, the schedule publish day, platform status, and the post copy.

So they're all correct.

And for the files and medias, just click on it to double check.

So all the generated social visuals for these carousel posts are being uploaded as well.

So this is perfect.

So imagine you're collaborating with a team using this Notion Calendar.

They can just open the calendar and everything they need is already there.

So they can review and mark comments Now, if that's one thing that is up the most time and budget on any marketing team is ads creative, it is a whole different challenge.

You need a creative brief, a visual directions, and enough variations to actually test what would be working.

So let me show you how the system can also handle that from campaign to creative brief to scaling and variations.

For our ad creative system.

So first we'll create the style library.

So similar to the carousel library we just did.

This library will focus more on the ad creative side, so it contains different style of the high performing reference, including static ads or carousel post ads, as well as the Ad creative style guide.

And then under the ads folder, so we have the input folder where we will include any reference images.

So this can be your product images, competitor reference, or maybe studio backgrounds.. So for this case, I have include two product images where I

backgrounds.. So for this case, I have include two product images where I want the model to reference explicitly.

So back to our content planner.

So under the Ad tab, we have the campaign name, the products and offers, the goals for this campaign, the platform, etc. So first we need to generate the creative briefs.

So I will use this prompt to load the content planner and generate a brief for each campaign.

So you can see it's spinning up parallel agents to handle all five at once, using the ad creative brief skill.

And just like that five brief are done quickly.

So each one follow the prebuilt template with campaign overview target audience, key message hook variations, creative directions, all structured and consistent.

And you can always tweak the template to fit your case.

And everything is saved inside this output folder, organized by campaign name, which is perfect.

So that means everything is ready.

We are going to generate the ad creatives, so again, back to cowork.

Use this prompt to ask it, to read through all creative briefs, to understand the requirements, and then generate the ad creatives.

So now you can see it's starting the ad creative style guide, calling the creative designer skills, setting up agents to generate the five sets of creatives.

So note by default, I have set the system to always generate five ad creatives for each design.

So this is why you can see finally we will have 25 ad creatives generated across the five campaigns.

And the best thing is they're all being saved inside the same campaign folder along with the brief, and they all look pretty impressive, so like this one where use the Grazing Cups reference image.

So you can definitely tell Nano Banana have used it across these five ads variations and what I really like is that they just have different vibes and moods to give you lots of inspiration.

And for other campaigns they are also well executed with different moods, matching the campaign theme and this one about the pumpkin season campaign.

Also using our reference image as the anchor, which is nicely done.

Now I would say though, these are still not production ready, so there's still things that needs to be fixed, so like the logo and the tag line.

But these will give you a loss of inspiration on how your products or service would look like in real campaign settings.

So across different formats and placement.

And I suggest to always include more of your reference image, So the final output quality will be even closer to what you want.

So far we have covered social carousels and Ads, but there is one channel that compounds over time more than any other that is organic search.

But the challenge is as your content takes the longest to produce.

So now we are going to batch, create full blog posts from content briefs, generate how images and push them straight to your website as draft.

So for this blog writing system, first we set up the blog content brief template.

And together with the AI search blog writing skills, we have already created different sets of content briefs that sits inside this output folder under SEO.

So it contains our topic and keyword strategy, content angles, structure and formats, AI search optimization notes, so we can standardize the whole process.

And now the key is to include them inside our content planner, which has the target topic content angle, as well as the content brief file path and status.

So now on cowork, use this prompt to ask it to read the content brief generated and then create a blog post draft and then also generate the feature header image for the blog.

So now it will use the AI search blog writing skills for the task, reading all the content briefs and spin up parallel block writing agents.

So once it is finished, it will kick off the header image generation task for each blog.

So the total process take around 15 minutes.

So each draft follows the content brief with table of contents key takeaways, bullet points, and structure formatting.

And what's great is that it also generate the blog header image so you can use for that article.

So each matching the blog content.

But we are not done yet.

So we are also pushing these article draft straight to the website, which is a WordPress in this case.

So for this purpose we need a WebPress MCP tool.

And since we are also uploading the images through the external API call, we will also use cloud code again.

So on cloud code desktop, I asked it to read through each post draft and push them to WebPress with the header image, creating each one as a new draft.

So it will load the content planner.

Pull the to-do list for all the seven posts and work through them one by one.

A quick tip is if you ever hit context window limit, you can use the built-in slash command compact to free up context space.

Or even better, split them up into separate Claude code sessions.

And now all the seven posts are created as draft on the website.

So we can double check from the back end.

So each one has the feature image that Claude upload the exact content with formatting optimized for AI search and it is exactly the version we see from the local draft.

But still, I highly recommend to have a human review first enhance it with your own insights, adding unique data graphics before posting But the time saving is huge, so you can just initiate the same pipeline on schedule.

So you can always have fresh content draft waiting for your posting.

If you enjoy this video so far, please give me a thumbs up.

I really appreciate that.

I also invite you to join our Growth community.

We are now running a workshop series about building AI system for marketing, where we go much deeper from context setup, research workflows, all the way to building a complete system.

You can find the link below to join.

I hope to see you there and join all the AI native marketers.

And before we go, if you want to learn how to create your first Claude skills for marketing, also watch these two videos.

I will see you next time.