0:02 Every month, AI agents on cloud code,
0:03 co-work, and codecs are getting better
0:05 at reasoning, writing code, and
0:07 navigating software. But for AI agents
0:09 to actually become our main interface
0:11 for work and doing work autonomously,
0:13 there's still one missing layer:
0:15 context. So, in this video, I'll show
0:17 you how I built a second brain in
0:20 Obsidian that plugs into co-work, or any
0:22 AI agent I use, giving them persistent
0:24 context and memory around me and my
0:26 business. I'll cover the five big
0:28 advantages of this setup and show you
0:30 why this might have huge implications
0:32 for how we work and how businesses are
0:33 run. And I'll show you how it actually
0:36 works and an easy way for you to set it
0:38 up and get started with this today. Now,
0:39 before showing you how this works, how
0:41 to set it up, and why I recommend
0:42 starting with this today, let me quickly
0:44 go over the five big advantages with
0:46 some examples of having a second brain.
0:48 Now, the first one is the most obvious
0:50 one, and it's persistent context. Right
0:52 now, most people use AI in isolated
0:54 conversations. You have to reexlain
0:56 everything in each chat about your
0:58 situation, your project, your workflows,
1:00 etc. And with a second brain, your AI
1:02 agent has persistent access to all of
1:05 this context and not just a few facts,
1:07 but detailed context around everything.
1:08 You can see in my Obsidian, I have
1:10 context saved around everything. My
1:12 business, my strategy, my projects, my
1:15 brand, my workflows, my team, uh about
1:17 myself, my meetings, literally
1:18 everything. And here in this graph view,
1:20 we can see the relationships between all
1:22 of these documents. for example, my AI
1:24 accelerator here and here's me. Now,
1:25 I'll show you later how this actually
1:26 works and how to set this up. But
1:29 because of this, I can now open a new
1:31 chat session with, for example, Cloud
1:33 Co-work, and ask something like, "What
1:35 should I focus on today? It's already
1:37 connected to my knowledge, Obsidian
1:38 Vault, how they call it." And you can
1:41 see Claude now pulls my context to give
1:42 me an answer. And because it pulled this
1:44 context, it now knows that my main
1:45 priority should be landing page copy
1:47 changes, recording the Obsidian video,
1:49 and organizing the Spain offsite in
1:52 April. I could even ask it things like,
1:54 "Write me a LinkedIn post based on AI
1:56 topics we discussed in our team meetings
1:59 this week. Use the LinkedIn instant uh
2:02 skill." It will now go through our team
2:04 meetings, see what AI topics we
2:06 discussed, and use a LinkedIn skill to
2:08 actually write it in my tone of voice,
2:09 then pulls the context from my second
2:12 brain, uses the skill, and then I'll put
2:13 a LinkedIn post according to the topic
2:14 we discussed, which is of course this
2:17 second brain topic. This week, my team
2:18 and I built something different. We
2:20 created a second brain for our entire
2:21 business. Now, this is just an example,
2:23 but you can see the power of this. And
2:24 with the new scheduled task feature,
2:26 this becomes even more powerful.
2:28 Secondly, besides co-work now always
2:30 being able to pull up to date and
2:31 complete context around me and my
2:33 business in any chat, it can also
2:35 directly update the context in my second
2:38 brain. So, any decision, any rule, any
2:40 project update I make in an AI chat, it
2:42 can log it directly back into my second
2:44 brain. For example, if I see something
2:45 in this LinkedIn skill that I don't
2:48 like, for example, never use um m dashes
2:51 when writing content for me, I can say
2:52 something like remember this in my
2:54 second brain or in my AI operating
2:57 system. You can see it now saved and
2:59 updated this in my second brain as a
3:00 role in the writing preferences. Now,
3:02 this is huge because it means the more
3:04 you and your team use AI to do tasks,
3:06 the more context is built, the more
3:08 guidelines it has and the better your AI
3:10 becomes for yourself and your entire
3:11 team. Thirdly, there's also a big
3:13 advantage of having this second brain
3:15 when building and using skills. Now, if
3:17 you don't know what skills are yet,
3:19 skills are basically saved instructions
3:21 for your AI agent on cloud code or cloud
3:23 co-work on how to do a specific process
3:25 or task. For example, in this LinkedIn
3:27 skill, it goes through the step-by-step
3:29 process of how to write a LinkedIn post
3:31 in my style and essentially allows these
3:33 agents to automate workflows just like I
3:35 showed you in the example. Now, I have a
3:36 full video covering skills if you're
3:38 still unfamiliar with it, which I'll
3:39 make sure to link in the description
3:40 below, too. But what changes with this
3:42 second brain? Skills usually have
3:45 reference files and context inside the
3:47 skill folder. For example, in this uh
3:49 LinkedIn skill, I have a pen profile
3:51 background document, a hook template
3:53 document, an ICP document, a LinkedIn
3:55 example document, and a voice
3:57 personality document. Now, we use these
3:58 reference files in skills of course to
4:00 get to better outputs, but it usually
4:02 takes a long time to provide all the
4:04 context every new skill needs to get to
4:06 a good output. And if you have all of
4:08 the relevant context around your ICP,
4:09 your tone of voice, your business
4:11 already in your second brain, it means
4:13 you can build good skills far faster.
4:15 This means you only need to lay out the
4:16 process and point it to the right
4:18 context in your second brain. And you
4:20 don't have to give it the exact same
4:22 context over and over again for each new
4:24 skill you're building. For example, I
4:25 have dozens of skills that share
4:27 reference files around my ICP, their
4:29 pain points, what my business does, etc.
4:30 As you can see, my newsletter writer
4:32 skill shares a lot of the same context
4:35 files as my LinkedIn one. So the new way
4:36 I'm building out my skills with this
4:39 second brain setup is not by adding this
4:41 extra context or reference files in the
4:43 skill itself but by just pointing my
4:45 skill to the right context in my second
4:47 brain. For example, this is my new
4:49 LinkedIn skill the LinkedIn instant and
4:51 in this case you can see that I only
4:53 have the skill MD only the process
4:56 instructions and there I direct it to
4:57 where I can find the specific context
5:00 files in my second brain. And this means
5:02 that any update I make to a reference
5:04 file in my second brain, for example, in
5:06 the ICP document, all of the skills that
5:08 use this ICP document are instantly
5:11 updated instead of me manually updating
5:13 dozens of skills myself. It also means
5:15 that this rule, for example, that I
5:17 added, uh, never use mashes when writing
5:20 content for me, is now updated as a rule
5:21 in my writing preferences, which is one
5:23 of the documents many of my content
5:25 writing skills point towards. So, my
5:27 newsletter skill is automatically
5:29 updated with this same rule. Now,
5:31 another huge advantage of this setup, of
5:33 course, is that it works across any AI
5:35 you use. So, your second brain in
5:37 Obsidian is really just a folder of
5:38 markdown files, which I'll explain in
5:40 more detail in a second. But that's all
5:42 it is. It's just a folder that I can now
5:44 give Clot Codework access to. I can give
5:46 Clot Code access to. I can give Codeex,
5:48 Anti-gravity, or any other AI agent
5:51 provider access to. For example, I can
5:53 go to the code tab or just use Cloud
5:55 Code in the terminal. give it access to
5:56 the same folder of my second brain the
5:59 Ben iOS and I can ask the same question
6:02 what should I focus on today
6:03 and as you can see clot code has access
6:05 to the same context as you can see
6:07 landing page copy Obsidian video and
6:09 Spain offsite and this even works across
6:11 different AI providers here I gave Codex
6:13 access to the same folder as the same
6:15 question as you can see landing page
6:17 copy YouTube production and Spain
6:19 offsite and then lastly which is huge if
6:20 you're a business is that this is not
6:22 only for yourself but can actually scale
6:24 across your entire team and business. Me
6:26 and my team now share this same second
6:28 brain with my business strategy, ICP
6:31 understanding, uh to voice references,
6:33 company goals, etc. So instantly, my
6:36 entire team's AI agents have access to
6:38 this context that make them far more
6:40 powerful and productive for my business.
6:42 For example, my entire team will always
6:43 have access to up-to-date strategy
6:45 documents, ICP documents that their
6:47 agents can instantly use. And with the
6:49 context and skills that I've built, I
6:51 can now let an engineer write my
6:53 LinkedIn post with an onpoint tone of
6:55 voice. In Obsidian, I can sync these
6:57 updates in my context across the entire
6:59 team here. And you can imagine that this
7:00 setup could completely change the way
7:02 you run a business, which I'll get to
7:03 later in this video. But let me first
7:05 explain how this actually works because
7:06 it might look a little bit overwhelming
7:08 and complicated, but it really is not.
7:10 So, what is Obsidian? All Obsidian
7:12 really is is just a visual overlay of a
7:15 folder and its files on your computer.
7:16 As you can see here, all these folders
7:18 that I'm seeing in Obsidian, contacts,
7:20 daily departments, intelligence,
7:22 onboarding, I also have available here
7:24 in the folder that Obsidian is connected
7:26 to. And all we really do is we point
7:28 cloud co-work, cloud code or codeex or
7:30 any other AI agent we use to the same
7:32 folder on your computer. So they can
7:34 directly read and write to the same
7:36 files you see in Obsidian. And that's
7:38 why Obsidian is a great app to do this
7:39 on because we don't have to sync
7:42 anything, use an API or an MCP or a
7:44 cloud-based software. It's all local. So
7:45 in coowwork for example, all I do is
7:48 give it access to that folder. But if we
7:49 have thousands of these context files,
7:51 how does our AI agent on clockwork or
7:53 clock code actually know what context to
7:55 use and how does it update it? Now
7:57 again, this is pretty simple. The way it
7:59 updates and retrieves the right data
8:02 from your Obsidian folder or vault is
8:04 through the cloud. MD file. Now the
8:06 cloud.md file is basically just an
8:08 instruction you give to your AI agent on
8:10 cloud co-work or cloud code on how to
8:13 navigate the second brain or the folder.
8:14 So if I ask coworker a question like
8:16 what did we talk about in our team
8:18 meeting yesterday? My AI agent first of
8:20 all knows that it needs more context to
8:22 answer this question. It then reads the
8:24 cloud MD file to understand where in my
8:27 Obsidian vault or folder it will find
8:29 more information around this. Then it
8:31 reads those specific documents like for
8:33 example yesterday's Firefly transcript
8:35 to answer the question back to the user.
8:37 So you can basically see that cloud.md
8:40 file as sort of like a system prompt or
8:42 an instruction layer that tells your AI
8:44 agent how the vault or your folder is
8:46 structured and where to retrieve and
8:48 save data. So as you can see in this
8:50 session with cloud co-work where we
8:52 wrote the LinkedIn post, it has access
8:53 to one file which is the instructions
8:55 for the cloud MD and this is basically
8:57 just instructions for the AI agent on
8:59 where to find specific information in
9:01 the second brain and where to save it.
9:02 So you can see how this system works,
9:06 the file structure, knowledge routing,
9:08 and this same clot.md file or the
9:09 instructions you also see here in the
9:11 Obsidian vault. Now, don't worry, I'll
9:12 show you later in this video exactly how
9:14 to get to this cloud MD and how to set
9:16 up these instructions easily. Now, you
9:17 might be asking, how's this actually
9:19 different from CL's built-in memory, and
9:21 why would we actually need Obsidian if
9:22 it's just a folder? Now, first of all,
9:25 CL's built-in memory is very limited and
9:27 is basically designed to remember the
9:29 most essential facts about you. It's
9:32 generally stored in one document. So the
9:33 difference really is the scope of the
9:35 context. My Obsidian Vault, as you can
9:38 see, has thousands of pieces of context.
9:40 And secondly, if Obsidian is just a
9:41 folder, why would we need Obsidian? The
9:43 short answer is you don't need it. Uh
9:44 you can set this up in a folder
9:46 yourself, too. But honestly, it's just a
9:48 nice way for you to visualize, organize,
9:51 navigate, uh search, and link your notes
9:53 and files together. my context and
9:54 knowledge sources over the last weeks
9:56 have been growing really fast and
9:58 honestly without Obsidian I wouldn't be
10:00 able to organize it the way uh that I
10:01 have right now through the graph view we
10:03 can also see the relationships between
10:05 all of these context files it
10:06 automatically makes these connections
10:08 between different documents or context
10:10 files uh for example in the brand
10:12 identity document you can see that we
10:14 have the voice positioning where links
10:16 to the ICP for our ideal customer
10:17 profile and the pain points of our
10:20 customer this is also what your AI agent
10:22 is able to navigate so for example if it
10:23 reads the brand guidelines and feels
10:25 like it needs more context around my
10:27 ICP, you can see this link or wiki link,
10:29 what they call it, and actually look up
10:31 this document to find more information.
10:33 And the nice thing about Obsidian is I
10:35 can really easily sync the uh updated
10:37 context across my team if you're going
10:38 to use this in a team setting. It's also
10:40 entirely free to use and download. So, I
10:42 just recommend using it. Now, before
10:43 showing you how to set it up, let me
10:45 zoom out for one second because I think
10:47 this setup has much bigger implications
10:49 than just some extra productivity. I
10:50 think it could entirely change the way
10:53 people work and businesses are run. So
10:54 there are multiple of these big
10:55 developments coming together right now
10:57 in AI. Uh everyone can see that these
10:59 L&Ms are becoming better at reasoning.
11:01 MCPs are getting better and now allow
11:03 them to efficiently navigate softwares
11:05 and the internet and skills plugins,
11:08 schedule tasks, etc. now allow you to
11:10 automate uh repetitive tasks fast and
11:12 easily. But the missing layer for those
11:14 AI agents was really context. And with a
11:16 setup like this, I think it will slowly
11:18 allow people and businesses to start
11:21 adopting an AI interface like cloud
11:22 co-work or cloud code as the main
11:24 interface to do their work instead of
11:26 hopping between 15 different softwares
11:27 all the time. But maybe more
11:29 importantly, I think this is the
11:31 development that will slowly allow AI
11:33 agents to start doing work autonomously
11:36 without our involvement. Personally, for
11:37 example, since I've really started using
11:39 co-work on a daily basis, I've been less
11:41 and less in my Gmail inbox. I've been
11:43 less on Google doing research. I'm
11:45 barely in my CRM anymore. And now with
11:47 these combination of MCPs, connectors,
11:50 and scheduled skills, I can now automate
11:51 end-to-end processes like email
11:53 follow-ups without my involvement. And
11:54 that's why I really believe you need to
11:56 start building this today because the
11:58 value of this setup isn't in the setup
12:00 itself. It's in the context that builds
12:02 over time. Every decision that gets
12:05 logged, every correction or role that
12:07 gets saved, every project that gets
12:09 documented, and every skill that gets
12:11 made, it all compounds. So the AI agent
12:14 you and your entire team have after 6
12:15 months of using this is far more
12:17 powerful than the one you start with on
12:19 day one. And if your competitor for
12:21 example starts 6 months after you,
12:23 they're not just behind on the tool,
12:25 they're behind on 6 month of
12:26 intelligence that makes the tool
12:28 actually perform far better for you. And
12:30 even when better models come out, and
12:32 they will, the same context just becomes
12:34 more powerful. So the context I think
12:36 will be your actual mode in the upcoming
12:38 months and years. So how do you actually
12:39 set this up for yourself or your
12:41 business? Now, the key thing to keep in
12:42 mind when you get started is that you
12:44 don't want to overoptimize. It might
12:45 look very overwhelming what I just
12:47 showed you in my own setup, but this
12:49 setup I started with just probably five
12:52 files a few weeks ago. This context will
12:54 grow very naturally the more you use AI.
12:56 You just want to sort of start very
12:58 simple and let the system evolve
12:59 naturally. The same is the case for the
13:01 file structure. It is important, but you
13:03 do want to start simple and let it
13:05 evolve naturally. There is really no
13:07 perfect file structure because it's
13:09 going to be highly context dependent. It
13:11 will depend on your context, your
13:13 business, your goals, and your projects.
13:15 Um, now that being said, there are two
13:17 file structures that I recommend and
13:19 I've seen work well as a starting file
13:21 structure. One for if you're running a
13:22 business uh and you want to use this
13:25 across your team and one if you uh want
13:26 to use this for yourself as a
13:28 professional or as a soloreneur. Now, go
13:30 over the file structure quickly so you
13:31 understand what's in each because it is
13:33 important. But we've also built a
13:35 plug-in you can use in your AI agent on
13:37 cloud co-work or cloud code to help you
13:38 set up and get to these starting
13:40 structures fast, which I'll show you in
13:41 a second how to use. And second, what
13:42 you want to keep in mind when I'm going
13:44 through this is that many of these files
13:47 are and can be created by your AI agent.
13:48 So don't get overwhelmed. You'll get
13:50 there naturally. So I'll cover the file
13:51 structure that I use for my business
13:53 setup and then I'll show you quickly the
13:54 personal setup which is basically the
13:56 same but with less files. So first we
13:58 have the context folder and this is
14:00 where you store general context around
14:01 who you are, your business, your
14:03 strategy, your team, your brand and it's
14:05 basically everything your AI agent needs
14:07 to understand about you and your
14:09 situation always. For example, in
14:10 context, I have information about my
14:12 team, strategy, stakeholders, pain
14:15 points, organization, operator, the ICP,
14:17 and the brand. Second, we have daily.
14:19 And this is basically where your AI
14:21 agent logs everything that happened each
14:23 day across your sessions, maybe across
14:25 your meetings. And this is probably the
14:27 most important one because it gives your
14:30 AI agent that continuity between
14:32 conversations. Then third we have
14:34 departments. Now this is if you run a
14:35 business uh you will of course have
14:37 different departments. For example in my
14:39 case community, content engineering
14:41 partnerships operations etc. And then in
14:43 the community folder for example we can
14:45 have SOPs around work that needs to be
14:47 done in my community. For example here
14:49 YouTube to community repurposing. The
14:51 fourth one is intelligence. And this is
14:53 a bit like the first one context but
14:55 much more detailed. And this is the
14:57 place where uh things like meeting
15:00 transcripts, decisions, uh competitor
15:02 research, market insights get stored
15:04 over time. Then we have onboarding. Um
15:06 here you can have SOPs around uh
15:08 onboarding new team members or even
15:10 clients. Then we have projects here. Now
15:12 projects will highly depend on uh your
15:14 context. For me, projects can be for
15:15 example different YouTube videos I'm
15:18 working on. So I can ideate and work on
15:20 scripting uh on one video between
15:22 different chats. Uh if you run an
15:23 agency, this can be a project for each
15:25 client you're managing. But this will be
15:27 highly context dependent. Then fifth, we
15:29 have resources. And resources is
15:31 basically anything reusable. So you can
15:34 imagine it like a library of prompts,
15:37 templates, frameworks, maybe content
15:39 output examples, good examples, things
15:40 like this. Then we have the skill
15:43 folder. Uh an important one uh where the
15:45 reference material of your skills live.
15:47 for example, your strategy docs, your
15:49 voice guides, your ICP descriptions, uh,
15:51 basically additional information that
15:53 your skills point to. So, you can see I
15:55 have all my skills here, uh, with the
15:56 reference files laid out. By the way, if
15:58 you want access to all of the skills
15:59 that me and my team are building out and
16:01 using, uh, you can also check out my AI
16:02 accelerator in the link in the
16:04 description. Then, lastly, we have here
16:06 tasks. And tasks can basically be uh,
16:08 to-do lists. And then we have teams with
16:10 more context uh, around each team
16:12 member's role and responsibilities in
16:14 your business. So your agent always has
16:16 context around anyone in your team. And
16:18 that's really it. And then at the root
16:20 here, you have the clot.md file, the
16:22 brain file, which is the instruction
16:24 layer that tells your AI agent how this
16:26 whole file system here works and how to
16:28 navigate it. And this will also appear
16:30 in the co-work section in their folder
16:31 instruction that I showed you before.
16:32 Now, if you're setting this up for
16:34 yourself uh as a solopreneur or maybe as
16:36 a professional, you can basically have
16:38 the same uh structure but a little bit
16:40 simpler. Uh so here I have an example of
16:42 the personal OS. It's basically the same
16:43 file structure uh without the
16:46 department, without the team section,
16:48 and without the onboarding. So, same
16:50 file structure, just a bit simpler. Now,
16:51 again, might look overwhelming, but I
16:53 started this a few weeks ago with just
16:54 five files, and this sort of naturally
16:56 grew, and a lot of this uh context has
16:58 been created by my AI agent. And the
17:00 plug-in we built is going to make this
17:02 process a lot easier and faster to do.
17:04 So, how do we set it up? You can just go
17:06 to Obsidian and download Obsidian for
17:07 free. Once you've done that, you'll land
17:09 on a screen like this. So you can just
17:11 go here to create new vault. For
17:14 example, I call it Benai test and then I
17:17 have to choose a folder. So I just
17:18 create a new folder and this is the
17:20 folder that you're going to point cowork
17:22 or cloth code or any AI agent you use
17:24 towards to access that same vault that
17:27 Obsidian visualizes for you. So we open
17:29 this one and click create. So we now
17:31 have an empty folder. Now if you're
17:33 going to set this up in cloud code, you
17:35 can also use Obsidian CLI which I'll
17:36 make sure to put in the description
17:38 below. that can help you get to a basic
17:40 generic setup a bit quicker. But as
17:42 said, we've built our own plug-in which
17:44 is available in my AI accelerator
17:46 together with all our other plugins and
17:47 skills to get you to that file structure
17:49 that I showed you before a lot quicker.
17:52 It helps you also populate uh your
17:54 essential context a lot faster. And this
17:56 plug-in will work across cloud co-work
17:57 or cloud code, wherever you want to use
17:59 it. So, if that's interesting to you,
18:00 you can check it out in my AI
18:01 accelerator in the link in description
18:03 below. We also have unlimited one-on-one
18:05 live tech help available if you want uh
18:06 some help setting up these things. We
18:08 also do AI workshops where we dive a lot
18:10 deeper into these setups and tools. So,
18:11 if that's interesting, definitely check
18:13 it out. Also, if you're a business and
18:14 you want me and my team to help you in a
18:16 more personalized way to set up a
18:18 business AI OS for your company, we're
18:20 now opening a few limited spots to help
18:22 businesses set this up. So, if you want
18:23 more information, you can also check the
18:25 link in the description below. In there,
18:27 you can find a link uh with the
18:29 marketplace of plugins and skills of my
18:31 accelerator. And from there, if you do
18:33 it in co-work, you can just go to
18:35 customize. Click here on the plus and
18:37 you click add marketplace and you add
18:39 the link that you find in accelerator.
18:41 Once you've done that and you go to
18:43 browse plugins, you'll find a tab here
18:46 at personal that says Benai skills. And
18:47 in there at the bottom, you'll find our
18:50 plugin, Ben Obsidian plugin. You can
18:53 install that. And now you'll see that in
18:55 your plugins, this one will appear. And
18:57 there we have the skills that help you
18:59 get to that setup in Obsidian a lot
19:00 faster. And all we do then is we point
19:02 co-work to the same folder that we just
19:04 set up in Obsidian. So in this case,
19:08 Beni test. We click always allow. And
19:10 now we uh use uh the plug-in and the
19:12 scale setup.
19:14 Right? We run this and it will walk you
19:16 through uh the setup on getting to this
19:18 file structure a lot faster. And I'll
19:20 start asking you some question to get
19:22 the essential context set up. So first
19:23 it asked me what kind of vault do I want
19:25 a business setup or a solopreneur setup.
19:28 So in this case I'll do uh solopreneur
19:30 just to show you as an example. So now
19:31 it's already created the initial folder
19:33 structure and the cloud MD with the
19:34 instructions on how to navigate this
19:36 type of folder structure. As you can see
19:38 now in obsidian we already have this
19:40 folder structure. Now most of these are
19:41 still empty of course because we haven't
19:43 given cloud any context. So the next
19:45 question is to really start giving it
19:47 context. Right? And that's what this
19:48 plug-in does, right? It it's going to
19:50 ask you some questions to populate your
19:53 initial context data set. now want to
19:54 spend probably half an hour to an hour
19:56 here to get your your initial setup have
19:59 the initial context data set and from
20:01 there it will naturally expand and I can
20:02 tell you if you start using AI more and
20:04 more in a couple of weeks you'll have a
20:07 very expanded data set of context that
20:09 really makes your AI far more powerful.
20:10 So a couple of important things to keep
20:12 in mind once you get to that initial
20:14 data set is every new task you start in
20:16 clock code you always want to point
20:19 towards that same uh folder. Second,
20:21 when there are things that you want your
20:23 AI or your second brain to remember,
20:25 clearly tell cloud co-work to remember
20:27 this in your second brain or whatever
20:29 you call your folder. If you can ideally
20:30 even point it to the specific file it
20:32 has to save that role to. Thirdly, if
20:34 you see it have issues navigating the
20:36 folder structure, tell it to update the
20:39 cloud MD, which you can also do yourself
20:41 because remember this is the bridge to
20:43 point it to the right direction. So, I
20:44 can add in rules all the way at the end
20:46 here too. You can see I've already added
20:49 some rules on how it should navigate the
20:51 folder structure. And then lastly, if
20:53 you're going to build skills, I highly
20:55 recommend to take a new approach now and
20:57 instead of embedding reference files
20:59 into the skill here in Cloud Co-work or
21:01 in Cloud Code, save the reference files
21:04 in your second brain and let the skill
21:06 point towards the right folders. You can
21:07 also adapt your old skills by just
21:09 telling claude. For example, here I
21:10 said, can you adapt my LinkedIn skill
21:12 and create a new skill and gave it here
21:14 specific instruction instead of having
21:16 the reference files in the skill point
21:19 towards the files in the Ben AIOS to get
21:20 the additional info instead of having
21:23 them saved in the skill. And again, the
21:24 earlier you start with this, the more
21:26 powerful your AI agent is going to get
21:28 over the long term. So, highly recommend
21:29 to start soon with this. Now, that's it
21:31 for this video. Again, if you want
21:32 access to all of the plug-in skills that
21:34 me and my team are building out, plus
21:36 unlimited one-on-one live tech help and
21:38 AI workshops where we dive a lot deeper
21:39 into these tools, you can check out my
21:40 AI accelerator in the first link in the
21:42 description. And if you're a business
21:43 that wants a little bit more
21:44 personalized help in setting this up for
21:46 your business, you can check the second
21:47 link in the description below. Thank you
21:48 so much for watching. If you want to
21:50 learn more about cloth co-work uh
21:52 skills, plugins, etc., You can also
