21:03Build Great Products · Tutorial
Claude Cowork Just Changed the Game
Chris Ashby 5-tip walkthrough of Claude Cowork — Claude Code with a GUI — ending in a live 15-minute app business plan built end-to-end.
January 13thThe argument in one line.
Claude Code's new mods feature isn't just for developer tooling: a few plain-language prompts turn it into a self-critique reviewer, a context-degradation warning bar, and a proactive feature-idea generator that any builder can install in minutes.
Read if. Skip if.
READ IF YOU ARE…
- You use Claude Code for strategy, brand, or business work and want it to flag its own blind spots instead of you taking every answer at face value.
- You've lost work quality to a degraded context window and want a built-in warning before responses get worse, with a one-click handoff into a fresh session.
- You're building a product with Claude Code and want proactive feature suggestions pulled from your own project, not just answers to questions you remembered to ask.
SKIP IF…
- You're looking for Anthropic-built, official mods — everything shown here is custom, prompt-built by the creator himself.
- You don't use Claude Code's desktop app, since every mod shown renders as a panel inside that specific UI.
The full version, fast.
Claude Code's new mods feature lets anyone extend the desktop app's UI with custom panels, and the demos Anthropic shipped are aimed squarely at developers. This video argues the real opportunity is workflow mods for any kind of Claude user: a self-critique panel that reviews Claude's own response for strengths and flaws before you act on it, a context-handoff bar that warns when a session's quality is degrading and offers a one-click new session with an auto-written handoff, and a feature-finder panel that reads your current project and proactively suggests three features to build next, each with a one-click 'build it' button. Every mod was built from a single one- or two-sentence prompt in roughly ten minutes.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 03:10
01 · Why mods are bigger than developer dashboards
Introduces Claude Code mods and the two official examples (Token Weather, Replay Theater), then argues the real opportunity is building mods for repeated non-developer workflows.
03:10 – 06:43
02 · Demo 1: Self-Critique mod
Built from one prompt, a side panel reviews Claude's own response and returns a verdict, strengths, flaws, and a suggested fix, so answers aren't taken at face value.
06:43 – 08:45
03 · Demo 2: Context-degradation handoff bar
A bar above the input field warns when context quality is degrading and offers a one-click new session with an auto-written handoff.
08:45 – 11:11
04 · Demo 3: Feature Finder mod
A globally-installed panel reads the current project and proposes three features to build, each with a title, description, reasoning, and a one-click build button.
11:11 – 13:43
05 · Bonus: Product OS progress tracker, and close
A mod built from one two-sentence prompt tracks percentage progress through the creator's four-phase Product OS system, then the video closes with a community CTA.
Lines worth screenshotting.
- Claude Code mods can extend the desktop app's UI for any task you repeat often, not just developer-facing dashboards.
- A self-critique mod that reviews Claude's own response for strengths and flaws took about ten minutes to build from one prompt.
- Reading Claude's output verbatim risks treating every answer as correct, when a built-in critique pass can catch overstated or one-sided claims.
- A raw context-used percentage isn't actionable on its own; what matters is a mod that tells you what decision to make with that number.
- A context-degradation bar can trigger a one-click new session with an auto-written handoff, removing the need to manually summarize progress.
- A 'feature finder' mod reads the current project and proactively suggests three features to build, each with a description, a reason, and a one-click build button.
- The best mods come from naming a task you already do over and over inside Claude, then turning that repeated action into a UI element.
- A single two-sentence prompt was enough to generate the Product OS progress tracker mod, which shows percentage complete across four build phases.
- Mods can be installed per-project or globally, so one mod you build becomes reusable infrastructure across every project in Claude Code.
Build mods for what you repeat
MOD STRATEGY
The real value in Claude Code mods isn't flashy developer dashboards, it's turning any task you already do repeatedly inside Claude into a one-click UI panel.
- A self-critique panel that reviews Claude's own response for strengths and flaws turns a single prompt into a standing check against taking answers at face value.
- Raw context-used percentages aren't actionable by themselves; a mod is more useful when it tells you what decision to make, like starting a fresh session before quality drops.
- A context-degradation bar can auto-write a handoff summary and launch a new session in one click, removing the manual work of re-briefing Claude.
- A feature-finder panel that reads the current project and proposes three features with a reason and a one-click build button turns planning into execution.
- Picking a mod to build starts with naming a task you already do over and over in Claude, not with imagining a developer-style dashboard.
- Every mod in the video came from a one- or two-sentence prompt and took roughly ten minutes to build, which lowers the bar to try one yourself.
- Mods install per-project or globally, so one well-built mod becomes reusable infrastructure across every project in Claude Code.
Terms worth knowing.
- Claude Code mod
- A custom panel or UI extension added to Claude Code's desktop app, created from a plain-language prompt rather than written as traditional code.
- Context degradation
- The drop in response quality that happens as a Claude session fills up its context window with accumulated conversation history.
- Handoff
- A written summary Claude generates of a session's progress so a new session can continue the same work without losing context.
- Product OS
- The creator's own system and Claude Code plugin for building a product end-to-end across four phases: define, design, develop, and distribute.
Things they pointed at.
Lines you could clip.
05:03
“think about the things that you do over and over and over again in Claude or in Claude code and build mods to actually do that inside of the UI that is going to help you to use Claude in a better way”
06:48
“I think a really good way to think about cloud code mods is to think about them from a user experience point of view, what decisions are you going to be making based on that amount of context being filled”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
Claude Code just released a new feature called mods which allows you to add custom plugins and UI to Claude Code in the desktop app but the trouble is most of the examples that you've seen online are all targeted at developers and yes Claude Code is a developer tool but I think the capabilities extend far beyond that and so in this video I want to show you three practical use cases for Claude Code mods outside of just development.
And these are use cases that nobody else is talking about. They're going to be super helpful for working with Cloud Code. I've even included a little bonus one at the end to show you how you can use it with my system for building products, Product OS.
Now I'm going to put the prompts I use to build all of these in the description down below, although they are super, super easy. If you don't know me, my name is Chris. And for the last 15 years, I've been designing apps and advising startups on product and design.
And I've helped hundreds of people to build and launch real products inside of my community over in school. And with that said, let's jump straight into it. ClaudeCodeMods and show you exactly what you can do.
So if you haven't heard about ClaudeCodeMods, basically ClaudeCodeMods are a way of extending the functionality of ClaudeCode. Now I can take you through all of the documentation here and explain exactly how it works and what it does, but I'm not going to do that. I want to just show you some examples that Claude Code gives for how you can build mods that extend functionality on top of Claude Code in the desktop app.
So the first example that they give is one called token weather, where basically you show a bar above your input field that shows you the amount of context that you've used and basically what the weather is like based on the amount of context. So if you have worse weather, you are used up more of the context basically here.
you can see here inside and this is the terminal version of Clawed Code here you can see here it says either clear for the weather forecast or showers if you use up a little bit more context or storm if you've used up even more context than that. Another example of a mod in Claude Code is called Replay Theater. Now, again, this is an example inside the terminal, and it's got this right -hand panel here, which allows you to replay the different steps through the thing that you've built in Claude Code.
But I actually want to show you Claude Code mods inside of the Claude Code desktop app, because I think this is where it gets super interesting. And you can add a actually nice UI layer on top of the functionality that you already have inside of Claude Code. So you can see here another example as well of a mod.
inside of Claude Code and this is one that is in the desktop app but this basically shows progress through a release pipeline and again this is more of a developer use case here but you can see how you can start to extend the functionality of the UI by adding these kind of custom widgets inside the UI to help you understand things inside of your different projects and you can build these on a per project or a global basis as well so you can kind of create a mod and then install it globally so that you can access it from inside of any project or you can set it up where you just install it plugin kind of on a per project basis and have that as a custom mod for a specific project.
And so now I've gone through roughly what Cloud Code mods is, I want to actually jump into Cloud Code and show you an example of some mods that I have built that I think provide a bit more functionality beyond these kind of developer specific use cases. And this is actually where I think it starts to get really, really interesting.
So I'm now inside of ClaudeCode here, and you can see on the bottom left, I've got the four mods that I have created. Now, the first mod that I want to show you is a self -critique ClaudeCode mod. Now, I built this just...
based off of literally a simple prompt here, and I'll put the prompts down below. So help me build a self -critique Claude code mod. I want the mod to review the Claude code message response and critique itself with a summary of flaws and strengths.
So basically creating a mod that allows me to critique Claude's own, basically allows Claude to critique Claude's own output and give it kind of strengths and weaknesses and ideas of how you basically might interpret that message. Now, this is something that I do when I ask Claude to kind of give itself a critique anytime that I'm doing strategic work or thinking about like how I want to build and develop my business over the next few years.
I also do this whenever I'm doing anything to do with brand or positioning for any sort of product and it just helps me to get a better read on the output that Claude is actually giving me back rather than reading into it verbatim because I think it's really easy to take these responses that Claude gives you and just accept them that they're you know they are exactly right and you know it gives you all the relevant information where I don't think it always does and it's good to have it kind of critiquing itself as it goes to kind of give you a better view of like how accurate or correct it actually is and whether or not your kind of opinion is being swayed in any specific direction.
So you can see here I've typed forward slash critique in this input bar down below now that the plugin is built and you can see you've got a bunch of options there as well that Claude has created so what I did was basically to build this plugin I'll load the critique plugin here and it will load this um load this panel on the right hand side and this is what I mean by Claude mods being able to build these custom UI elements inside of Claude so this is not native this self -critique is not native to Claude code this is just something that I have built inside of Claude code literally in about 10 minutes here And it's based on something that I do over and over and over again.
And so that is what I would also recommend as well is think about the things that you do over and over again in Claude or in Claude code and build mods to actually do that inside of the UI that is going to help you to use Claude in a better way. So you can see here you've got options inside this self -critique view to rerun.
It turns auto to on or off. So you can automatically run a self -critique if you're doing a specific like strategy deep dive or research session. And you can choose the model that you want to use.
here as well so either we use the same model as the session or we can choose to go between opus on it or haiku for this verdict and it says verdict here is mixed and what it gives you is it gives you a summary it gives you the strengths and that have been presented to you and it gives you the flaws in any arguments from Claude inside of this message back now this is just a message back basically about the mod that I've built here so this is not massively useful but if I ran this against a strategy session or against a technical architecture that Claude Coders presented it would give me a lot more thorough understanding of that and a lot better ability to pick out where the strengths and weaknesses were in whatever Claude was proposing to me and so you see here it gives me the flaws and then proposes a fix And so we can literally kind of use those fixes and one thing that I might like to build into this in the future because you can continue to develop these mods as you're working on them as well is actually a run the fix or like make this fix kind of button in this UI so that I can just click like make the fix and then it would come back and prompt Claude to like actually make that fix based on the self -critique here.
So that's the first Claude code mod that I have built. The second one that I want to show you is something similar to the context weather report. i showed you initially but something that's a little bit more useful than that because i think just showing the status of your context it's not super useful like what percentage of context have you used and not used you already get that in this visual ui down in the bottom right here and you can click it and you get all this information okay great i've used that much context here's my session limit like what do i actually do with that information so i think a really good way to think about cloud code mods is to think about them from a user experience point of view what decisions are you going to be making based on that amount of context being filled.
So what I've created here is basically a mod that has a bar above the input bar that prompts me as the user to start a new session when the context starts to degrade with a button to start a new session in the same project. And then I also added in here, I think Claude actually suggested this, was to add in, if I run this now, so you can see this is the mod actually in action here.
So I created this bar that pops up above the input field here. it shows the percentage of context that you've used but it also shows when the quality starts to degrade in your session because of the amount of context that you've already filled up and it will recommend you to start a new session and it will allow you to do a new session with a handoff as well.
So if I say new session with handoff it will then write a handoff using Claude and give it to a new session in Claude so that I can continue working where I left off or you can just do a fresh new session and these have hotkeys associated with them as well so instead of just getting a basic view of the context okay great I've used a certain amount of context it's rainy whatever that might mean I'm actually getting a something that's useful here in that it actually allows me to just create a handoff and jump into a new session without actually writing that handoff or like creating a skill or doing that myself it's just built into this mod which is going to automatically show on to get over a certain threshold of my context and then the third mod that i want to show you here is a mod that i built called feature finder which basically as you're working on a project will proactively think of new feature ideas for you to build and give you basically a description of the feature and also why i should add it to the project now i think this is something that i really really liked from a lot of the old school like i say old school but from like last year the old school like ai app building platforms where they would like suggest something that maybe you actually didn't understand
about the project or didn't know to add maybe around like security or auditing or like error tracking those sorts of things where maybe you wouldn't actually have thought about that but it's good to get like the AI to recommend these things as you're building so that you can build in like a better call basically build a better quality product that's more robust that has more security affordances all of that sort of stuff so if I jump into a project here And I do a new session, say in this Squash Pad project, which is just like a canvas, design canvas tool that I've been building.
And go into, and I just type in forward slash feature ideas. It's going to pop up with this side panel here, which basically reads the project and thinks about what's to build next. And so now we've got these ideas for this Scratchpad app that I'm building, restore prototype versions from history, annotation aware revision prompts, and then prototype export and public embed.
And you can just click build it to build each of these features. And that will, if I click build it there, let's click build it on this one. so that will send a message in here that says build this feature restore prototype versions from history add a one click restore to restore this version and you can see it's actually got even more detail in here related to the convex schema that we're using the why we want to build this is reading these prototypes and actually building this in now and so this is going to build this into my project and we can click new ideas in this feature finders field and that's going to give us new ideas for our product as well i'm actually going to stop this thread because i don't I don't want to build that right now.
I just want to show you how it works. But this is a globally installed plugin that I can now use across any of my projects. And I literally gave it this two sentence prompt here.
I'd like to create a code code mod that offers feature suggestions based on the project I'm working on that would add value to the project. So just three things I should build with the description and apply a button and why and a button to build it from the mod UI. And that's literally it.
And then I can improve on these and iterate over time. Now the last one I want to show you is related to my system for building products with AI tools called Product OS. And you can learn a bit more about Product OS and how to build with it if you go over to my community.
But basically it's a full system to build your app from start to finish. But what I created for my own system, and Product OS is a plugin that you can install with Cloud Code, Codex or Cursor as well, is I created a plugin called Product OS. I created a mod called Product OS that you can run inside of any project that has Product OS that you can create super, super easily using Claude mods.
And it just shows you the percentage of progress that you are through Product OS. So inside of Product OS, you have these four kind of key sections. You have define, which is about figuring out who your customer persona is, your pricing strategy, your product offer, design, which is about your brand identity, UX, writing, design system.
design prompts your onboarding flow develop which is about the prd and the roadmap and building the mvp and then doing a security audit and then distribution which is about goes and also deploying to customers as well and launching your product and then distribution is about go to market strategy high growth experiments and then logging your results and then scaling and automating that and it also says what's up next for you to do as well and i want to also add in here you can add in a button to actually just run the product or a skill that you've got coming up next as well And again, this was literally a mod that I built from writing this one sentence of a prompt.
I want to create a Cloud Code mod that shows me my progress through Product OS, the plugin, based on the documents I have completed. And it's literally just built that mod for me. And these...
These mods are going to make my workflow and mods are going to make your workflow so much quicker and easier inside of Clawed Code. They're going to add all this custom functionality for you and your projects, whatever you're building, and they can add functionality beyond just developer workflows as well. That is super, super powerful.
And so I hope you understand how you can design these mods inside of Cloud Code and how they can be helpful for your own work with AI as well. If you are building something with AI and you want to launch it, turn it into a real product, turn it into an AI business, and you want to get support on your build if you're stuck or you just don't know what to do next, then you can check out my community over at school .com forward slash AI apps and the link is in the description down below.
If you enjoyed the video and if you thought it was helpful, then don't forget to like and subscribe. Thank you very much for watching and I'll see you next time.
The Hook
The bait, then the rug-pull.
Anthropic shipped Claude Code mods and every example online is a developer dashboard. This video skips those and builds three mods for the work Claude users actually repeat: a self-critique panel, a context-degradation warning bar, and a feature-idea generator, each built from a single plain-language prompt.
Named ideas worth stealing.
11:11model
Product OS four-phase build system
- Define
- Design
- Develop
- Distribute
Define covers customer persona, pricing strategy, and offer. Design covers brand identity, UX, writing, and design system. Develop covers the PRD, roadmap, MVP build, and a security audit. Distribute covers go-to-market strategy, growth experiments, and scaling.
Steal forany mod or dashboard that needs to track progress through a multi-phase build process
How they asked for the click.
VERBAL ASK
13:23link
“check out my community over at school.com forward slash AI apps”
Single verbal CTA after the final demo wraps, paired with a like-and-subscribe ask; soft-sell, no discount or urgency language.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
