← Free learning

Free video guide · No account needed

Getting started with AI desktop agents

Learn your way around AI desktop agents, from setup and model choices to permissions, skills, connectors, and practical workflows. Watch this five-part class at your own pace.

PART 115:26

Getting started, app tour, and model choices

Transcript

Automatically generated transcript; may contain recognition errors. Timestamps use HH:MM:SS:FF (25 frames per second).

[00:00:00:00]So, okay.

[00:00:02:03]Official intro, I am Novae Caelum.

[00:00:06:19]I'm an author, I am a filmmaker.

[00:00:09:11]I use AI to supercharge things.

[00:00:13:05]I'm mostly a direct sales author,

[00:00:15:04]so I work heavily with Shopify.

[00:00:17:13]And when AI agents came around that could talk to Shopify,

[00:00:22:01]like that'll go way easier.

[00:00:24:16]I use AI for editing.

[00:00:27:18]And when I say I use AI, it's mostly the desktop agents.

[00:00:32:03]It is mostly things that can do stuff for me.

[00:00:39:15]I am using it to help me with AI filmmaking.

[00:00:43:24]I'm using it for editing.

[00:00:45:12]I'm using it for newsletters and emails and just all kinds of stuff.

[00:00:52:11]Go out and do research.

[00:00:54:02]So, I'm going to be showing you first what the desktop agents are,

[00:01:01:05]walking through the basics of where to get them, how to download them,

[00:01:05:02]what all the menus are inside of them,

[00:01:08:20]what skills are, what connectors are, MCPs,

[00:01:13:14]how to talk with them, how to work in a mindset where you are building systems

[00:01:19:02]and not just having a chat where we still have a few people coming in.

[00:01:25:07]So, like most people at this point know how to use chatbots,

[00:01:28:24]like ChatGPT or even Claude or Gemini or any of those.

[00:01:36:00]Chatbots, you just have a conversation.

[00:01:38:03]And the desktop agents are similar,

[00:01:41:00]but with the difference that they can do things for you.

[00:01:45:13]They can take over your computer.

[00:01:47:00]They can go onto websites and do things for you.

[00:01:50:05]They can make you apps.

[00:01:52:15]They can take large bits of information and synthesize them.

[00:01:58:09]They can do a whole lot of things.

[00:02:02:05]Now, with OpenAI's Astra, Astra can go into like almost any app on your computer

[00:02:08:23]and just take it over and start doing things, which is kind of insane.

[00:02:13:11]I haven't done this personally, but I know people

[00:02:17:00]who have been using it to format with Vellum for books.

[00:02:20:06]So, there's so much you can do with it.

[00:02:23:13]And if you are on the coding end of things, if you're interested in that,

[00:02:27:21]that is what they were initially built for to do coding,

[00:02:31:18]but then realizing that people can just do all of these other things.

[00:02:36:03]So, there's some of this that you can do like on ChatGPT online or Claude online.

[00:02:41:05]You can hook some connectors and skills and stuff to it now,

[00:02:44:21]but the power of the desktop ones is that it's working in your own files.

[00:02:51:02]And they used to be a little shaky with that, but they're a lot more stable now.

[00:02:57:11]So, we'll go over like how to use it safely and just precautions that you take,

[00:03:03:09]like backing everything up, everything that you put an agent on,

[00:03:07:14]make sure you have a backup somewhere.

[00:03:10:03]And I say that as someone who plays pretty fast and loose with this

[00:03:13:05]because you get a lot done when you give it more permissions,

[00:03:19:07]but there is a little bit more risk to that, but you just back everything up.

[00:03:23:18]So, we're going to look at, I'm going to share my screen here.

[00:03:28:20]So, the main desktop agents are like everyone knows ChatGPT at this point.

[00:03:37:14]Everyone knows ChatGPT and Claude.

[00:03:42:15]We have ChatGPT and Claude.

[00:03:46:19]Oh, I guess we have a new Opus today.

[00:03:51:00]ChatGPT and Claude, like we're familiar with this on the desktop.

[00:03:54:19]And okay, chatgpt.com slash download is where you get the desktop app for that.

[00:04:00:23]And Claude, I believe, is something similar.

[00:04:06:08]Where you just basically search download for Claude.

[00:04:12:19]And they're actual apps that you will download.

[00:04:15:24]So, you will get that from here or through the ChatGPT.

[00:04:24:20]chatgpt.com slash download.

[00:04:27:03]When you download these, it'll be like, I've gotten, I've had these for a while.

[00:04:36:21]So, I'm going to walk you through the things.

[00:04:41:07]I might need to sign in again here.

[00:04:44:00]This one.

[00:04:46:23]I'll walk you through the things that are in these apps, but the onboarding might

[00:04:52:20]be a little bit different now than when I initially started.

[00:04:55:04]But usually, it'll walk you through just starting up depending on the app.

[00:05:06:07]Where your stuff is, maybe what your personalization preferences are.

[00:05:10:24]It's usually not much.

[00:05:12:00]It's very, very simple.

[00:05:15:07]So, within the apps themselves, we're going to look at ChatGPT first because I

[00:05:20:04]think that's the one more people need to be looking at right now.

[00:05:27:02]And I think that's the one more people will be using.

[00:05:30:00]There's different parts of this.

[00:05:32:04]There is the ChatGPT part, which is where you kind of do still chat.

[00:05:38:17]And you can do like work and stuff in here, but it is separate.

[00:05:44:20]And then there's the Codex part where Codex is more geared around coding.

[00:05:50:16]But this is the one I think that we need to use because

[00:05:55:08]this is the one that can just take things and run with it.

[00:05:59:03]So, there is ChatGPT in Codex.

[00:06:01:07]So, if you're just sort of casually chatting, you go to the ChatGPT area.

[00:06:06:02]If you want to get stuff done, you are going to want to be in Codex.

[00:06:09:18]And similarly, in Claude, they have this separate.

[00:06:15:03]They have the Claude chat area up here in the upper left here.

[00:06:21:05]This is chat and Cowork, which is a little bit confusing

[00:06:26:15]and I actually think they're going to be combining these.

[00:06:29:07]So, the chat, again, is just where you if you just all if

[00:06:32:09]all you want to do is have a chat, this is where you go.

[00:06:35:09]If you want to work with stuff on your computer,

[00:06:41:07]with your apps, with things that you have connected,

[00:06:43:19]you can use Cowork.

[00:06:45:06]If you want to do things more technical or kind of

[00:06:48:22]more rigorous up here, you flip over to Claude code.

[00:06:54:02]I mostly work in Claude code and you don't have to do code in Claude code, but it

[00:07:02:03]has some more advanced versioning controls.

[00:07:05:15]Like, even if you're not even if you're not doing

[00:07:11:05]code, there's some features in here that are very useful.

[00:07:16:10]So, I'm mostly going to be showing Cowork or code.

[00:07:23:03]And the difference also is that Cowork is it has a sandbox environment.

[00:07:28:20]So, it's a little bit safer to work on your computer if

[00:07:31:16]you don't really know what you're doing with any code.

[00:07:38:08]So, if you're completely new to all this, you probably want to work in Cowork if

[00:07:43:07]you are in Claude and more advanced to go into Claude code.

[00:07:49:10]So, within all of this, I'm going to hop back to ChatGPT.

[00:07:55:10]Within all of this in the settings, the number

[00:08:00:21]one thing you need to keep an eye on is usage.

[00:08:05:20]Both of these apps handle it a little differently.

[00:08:08:14]The ChatGPT app, you have a weekly limit and it's based on tokens.

[00:08:17:12]And tokens, most of you probably know what tokens

[00:08:20:21]are, but they're the units that it's not quite a word.

[00:08:24:17]It's a couple of characters.

[00:08:27:20]It's what is broken into for the information to be sent and come back from AI.

[00:08:34:04]So, you get X amount of tokens per week.

[00:08:37:01]You don't really have to worry about how many.

[00:08:39:01]You just need to watch how much is on your usage.

[00:08:43:17]When you're doing heavy agent tasks, you're going to

[00:08:46:03]see tokens slurp down more than if you're just chatting.

[00:08:51:10]Because what's happening with heavy agent tasks is it's spinning up subagents.

[00:08:56:22]And the subagents are going off and doing things and

[00:08:59:20]it's calling tools and it's doing all of these things.

[00:09:03:09]Versus when you're chatting, it's just kind of

[00:09:05:07]chatting and maybe researching in the background.

[00:09:08:06]So, chatting kind of spins subagents too, but it's much

[00:09:11:23]more when you're working with the tool calling agents.

[00:09:19:01]So, you just keep an eye on your usage.

[00:09:22:02]ChatGPT app, they often give you resets.

[00:09:25:16]If you get really low on your usage, if you're down to

[00:09:28:15]like 1%, you can use a reset and it'll pop it back up.

[00:09:33:06]If you use a reset, usually it resets your time to a whole week again.

[00:09:43:10]If you are in Claude, Claude works a little differently.

[00:09:47:06]If you go into the settings here and usage is where this is.

[00:09:52:09]Claude works a little differently where there is an all model usage.

[00:09:56:04]There is Fable.

[00:09:58:00]If you're on a max plan, which is $100 or $200 plan, you will have this bar too.

[00:10:06:09]This apparently Claude now gives us resets.

[00:10:12:09]That's nice.

[00:10:13:04]That's new.

[00:10:15:24]So, within Claude, you have a current session.

[00:10:21:11]This is how long?

[00:10:23:06]This is about a five hour window.

[00:10:24:20]If you hit this, you will have to stop even if you still have usage for the week.

[00:10:30:07]This is the weekly usage and then this is the Fable usage.

[00:10:34:12]Fable usage adds to the all models, but Fable is,

[00:10:40:01]you can use up to half of your weekly usage with Fable.

[00:10:43:13]That's how that works.

[00:10:45:20]Fable is the big guns.

[00:10:49:08]Fable is not what you want for your everyday tasks for the most part.

[00:10:55:12]Fable is very, very smart.

[00:10:56:17]Like if you are asking about a grocery list, do not use Fable because it will just

[00:11:01:08]slurp up your usage and he's too smart for that.

[00:11:06:08]Not necessary.

[00:11:09:20]And same thing in ChatGPT, you have different models.

[00:11:16:20]So, if you're in here, I'm in the Codex version of the chat

[00:11:22:17]GPT app and there's all these little controls around here.

[00:11:28:02]You will have the model picker here, which Astra, it's funny, their models kind of

[00:11:34:20]go from like big cosmic to little cosmic like moon, earth, sun, galaxy.

[00:11:44:24]Let's run in.

[00:11:46:14]So Astra is their big model.

[00:11:48:06]Astra is a beast.

[00:11:49:24]Astra is the best model out there right now.

[00:11:52:15]Just hands down the best model out there.

[00:11:55:04]Astra can do most things for you.

[00:11:59:16]Sol is also very, very good.

[00:12:02:07]Terra and Luna are going to be like, if you just want to have a chat and

[00:12:06:12]conversation, Luna is, it's their least model.

[00:12:12:18]You're not going to want to be letting Luna take over your computer.

[00:12:17:22]Astra and Sol, you can let take over your computer, but

[00:12:21:23]these ones are just more like the small task and chat models.

[00:12:25:12]That said, Astra can spin up.

[00:12:30:00]When you're working with Astra, Astra can

[00:12:31:20]orchestrate lesser models to do other tasks.

[00:12:35:13]Same thing in Claude.

[00:12:37:18]In Claude, let me go back to Cowork.

[00:12:41:08]This is, yeah, we're, we're, Claude could Claude Cowork.

[00:12:45:01]So if we're in Cowork, this is the model picker here.

[00:12:49:05]Fable is the top one.

[00:12:51:07]This is the big guns.

[00:12:53:14]This is what you want for your crunchy tasks.

[00:12:56:05]Opus 5.5 is brand new, like as of today, I think.

[00:13:02:01]And Opus, it tends to be the workhorse for like most things.

[00:13:08:18]But Opus can, Opus kind of spins out on more complex things.

[00:13:13:09]So that's where you bump to Fable.

[00:13:15:20]Sonnet, I rarely use Sonnet just because Opus is very adequate.

[00:13:22:02]But if you are on a lesser token plan, if you're like

[00:13:25:20]on a $20 a month plan, Sonnet can handle a lot of stuff.

[00:13:29:10]Like Sonnet can go through email.

[00:13:31:07]Sonnet can handle a lot of like data entry and information tasks.

[00:13:37:19]So, and I almost never use Haiku.

[00:13:40:02]Haiku is just, it's their least model.

[00:13:44:20]With all of these, there are effort levels.

[00:13:48:13]I tend to keep to high or medium for most things.

[00:13:53:07]This is how much the model thinks and it uses more tokens when it thinks harder.

[00:14:02:11]Sometimes there is a trade-off where if you have it on medium and you give it a

[00:14:06:21]really tough task, it might take longer to think it through.

[00:14:11:04]And therefore, it will use more tokens anyway than it might have on high.

[00:14:14:12]So, we have, this is, it's how fast the models, or how

[00:14:22:03]much the models think to get done what you need to get done.

[00:14:26:06]So, I mostly keep it on medium and high.

[00:14:30:15]And I tend to just keep it high on default.

[00:14:33:20]Bump it down to medium if I'm low on tokens.

[00:14:36:18]But sometimes when it's on medium, it actually takes more

[00:14:40:05]tokens to think about a really, really tough task than high.

[00:14:44:24]It just kind of depends on the task.

[00:14:46:16]I almost never put it on extra and max, unless I am trying

[00:14:50:11]to get it to think through something very, very difficult.

[00:14:53:19]In which case, I will get it to think it

[00:14:55:09]through, make a plan, and then bump it back down.

[00:14:58:14]Same thing in the ChatGPT.

[00:15:01:11]This is the slider in the ChatGPT.

[00:15:05:07]Please don't put this on this, on Ultra, and think that you are going to get the

[00:15:10:01]best results because you might, but it'll just chew through your account.

[00:15:15:15]And almost everything high is adequate for

[00:15:19:06]almost everything, and even medium on many cases.

[00:15:22:09]So, I tend to just leave things on high.

Notes

Topic times use MM:SS within this episode. Adaptable prompts are templates based on the lesson, not verbatim quotations. App names, options, and capabilities reflect the recording and may change.

Desktop agents: what they add — 01:25

A chatbot primarily responds in a conversation. A desktop agent can also take actions with files, applications, browsers, and connected services. Useful tasks include organizing information, researching, drafting newsletters, preparing marketing assets, and editing video.

The main advantage of a desktop workflow is access to your working materials. Instead of copying everything into each chat, point the agent to the files and context needed for the task.

Before granting file access: make recoverable copies of important material. Broader access gives an agent more room to work, but also more room to make an unwanted change.

Choose the workspace for the job — 05:15

Choose the workspace for the job — 05:15
Workspace shown in the classMain use
ChatGPT chatConversation, questions, and discussing ideas.
CodexTasks that involve tools, files, applications, and carrying work through multiple steps. Coding is one use, not the only use.
Claude ChatConversational work.
Claude CoworkWorking with files and connected applications in a more bounded environment; a starting point for less technical users.
Claude CodeMore technical or involved work, including workflows that benefit from version controls. It can be used for non-coding tasks too.

The desktop apps have their own onboarding. Menu locations and the separation between chat, work, and coding areas can change, so focus on which environment has the access and tools your task needs.

Watch usage, not just the number of messages — 07:55

  • Tokens are units of information processed by the model. A message is not a fixed amount of usage.
  • Tool calls, multiple steps, and work distributed to subagents can use much more of an allowance than a short conversation.
  • Check the usage panel before starting a large job. A short session limit and a weekly limit can both matter; available weekly capacity does not necessarily mean the current session can continue.
  • Check the reset time and any reset controls available to your account. Do not assume every plan has the same allowances.

Match the model to the task — 10:45

Use a model with enough capability for the work instead of automatically choosing the most expensive option.

Match the model to the task — 10:45
Type of workSelection principle
Short questions, simple lists, casual chatA lighter model may be sufficient.
Routine information work and data entryStart with an everyday model that can handle the tools and steps involved.
Complex research, difficult planning, or extensive computer workUse a stronger model and allow more reasoning.

Models discussed in the class include Astra and Sol for substantial agent work; Terra and Luna for lighter work; and Fable, Opus, Sonnet, and Haiku on the Claude side. These are the options and preferences discussed in the recording, not a permanent ranking.

Set reasoning effort deliberately — 13:44

Model choice and reasoning effort are separate controls. Medium or high effort is the everyday range used in the class; the highest settings are reserved for unusually difficult problems.

Lower effort does not always mean lower total usage. A difficult task may take longer or require more attempts at a lower setting. One approach is to use higher effort to reason through a difficult plan, then reduce it for the resulting routine work.

Starting checklist: choose the workspace → back up the relevant files → select an appropriate model → set effort → check usage → give the agent a concrete task and the materials it needs.

PART 216:14

Permissions, skills, and reusable context

Transcript

Automatically generated transcript; may contain recognition errors. Timestamps use HH:MM:SS:FF (25 frames per second).

[00:00:00:00]There's also this thing, the permissions box.

[00:00:05:00]The safest way to use these agents is to have it ask for approval.

[00:00:09:14]However, it'll ask you every two seconds, and that is kind of defeating the point

[00:00:15:11]of letting an agent go off and do things for you if you

[00:00:17:12]have to sit there and hit approve, approve, approve, approve.

[00:00:21:03]So, I tend to put these on approve for me.

[00:00:24:23]It will approve most things on its own, and every now and

[00:00:28:17]then it'll come up with something that it decides is unsafe,

[00:00:31:05]and then it'll surface it to you and say, "Do you give me permission?"

[00:00:36:14]So, this is probably what you want on this one.

[00:00:40:07]Approve for me.

[00:00:41:19]In Claude, it's similar.

[00:00:45:11]Where's the permissions?

[00:00:49:09]Actually, I don't know where this is.

[00:00:51:02]They moved it.

[00:00:52:22]In Claude code, it is down here.

[00:00:55:13]Like, if you're in Claude code, it's this auto.

[00:00:59:11]There's manual, accept edits, plan, bypass permissions, and auto.

[00:01:04:05]Nine times out of ten, you want things on auto.

[00:01:06:21]It'll do the same thing like with chat GPT and handle permissions.

[00:01:13:02]I think...

[00:01:17:17]Okay, this is where it is here in Cowork.

[00:01:20:22]So, you have manually approve, automatically approve, skip all approvals.

[00:01:25:08]I have it on skip all approvals just because I

[00:01:27:09]hate hitting approve and I back my system up.

[00:01:31:01]But the safest one is automatically approve.

[00:01:35:04]It'll do it on its own, and then it will ask for anything that it thinks is unsafe.

[00:01:40:17]However, there's a lot that Claude thinks is unsafe

[00:01:43:15]that is fine, and it might just ask you a bazillion times.

[00:01:47:12]So, if you're getting fed up with that, you can put it on skip all approvals.

[00:01:51:12]The safety with this, though, is if you are doing things in your browser, if you

[00:01:57:17]have it going off and doing things for you in your browser,

[00:02:03:21]you can get what are called prompted injection attacks.

[00:02:06:18]And there's things that people can leave online

[00:02:12:24]maliciously that your agent can pick up and be like,

[00:02:15:24]"Oh, I will give you this information or that information."

[00:02:19:15]That is a possibility.

[00:02:22:02]If you are working on sites that you know, like

[00:02:26:07]if you're sending it off to research on Amazon,

[00:02:29:03]if you're sending it off to go look at your

[00:02:33:01]Gmail or something, that's less risky in that sense.

[00:02:37:15]The other thing with safety is when you give it

[00:02:41:17]permission to, you can have it work in a project folder.

[00:02:48:22]So, like, I could have it work in this Dropbox

[00:02:53:05]folder here, and then I will tell it to allow that.

[00:02:57:22]And then it will just work in this one folder with Cowork.

[00:03:02:08]But you can also, if you're in Cloud Code, you can

[00:03:08:02]give it the permission, like you can set a folder,

[00:03:10:15]and it'll see that folder, but it can also go

[00:03:14:16]off and look at everything else on your computer.

[00:03:17:05]I personally let it do that because I get a lot done quickly that way.

[00:03:22:19]However, back your stuff up because you want your whole computer backed up

[00:03:27:15]because there have been instances of agents that

[00:03:31:00]have gone, "Oops, I wiped your download folder,"

[00:03:33:24]or "Oops, I wiped your documents folder."

[00:03:37:16]So, if you give it permission to have the files all over your

[00:03:42:20]computer, just make sure you have an external drive back up,

[00:03:48:11]something backing up everything that is important

[00:03:51:18]because there is, it's not as big of a chance as it used to be,

[00:03:55:24]but there is still a chance that an agent could go,

[00:03:58:05]"Oops," and wipe something that you don't want it to wipe.

[00:04:02:00]So, just back things up, and if you're not

[00:04:08:10]comfortable with that, you don't have to.

[00:04:09:15]You can just give it one folder at a time, and

[00:04:13:24]there's a lot you can get done in one folder.

[00:04:16:06]Similar things with ChatGPT. I think in ChatGPT app,

[00:04:21:19]it's in the settings here, and I forget exactly where it is.

[00:04:31:07]It's in one of these. I think it might be in configuration.

[00:04:36:18]In configuration, you can tell it how much access you have, workspace, right,

[00:04:47:01]read-only, workspace, right, full access.

[00:04:50:02]So, I give it full access because I want it ripping

[00:04:52:23]across my files to get things done quickly for me.

[00:04:57:02]But the safer way to go is to do workspace, right,

[00:05:00:23]where it can only work in the folder that you give it.

[00:05:04:21]So, similar thing here is the approval

[00:05:09:16]settings, and you can do that on the chat itself.

[00:05:15:04]So, other things that might be important to know, we're going to look at...

[00:05:25:02]So, okay, let me back up.

[00:05:28:14]One of the things, one of the reasons the desktop agents are super powerful is that

[00:05:35:09]when you are giving them tasks, you can

[00:05:39:11]give them repeated instructions called skills.

[00:05:43:04]And many of you will have heard of or worked

[00:05:47:06]with skills before. A few of you might have not.

[00:05:50:22]There are instruction files. It's a set of instruction files that gives the AI a

[00:05:59:18]repeated ability to perform a task in a same or similar way, very similar.

[00:06:07:05]So, when I am doing anything in here, this is one of my film chats for filmmaking.

[00:06:15:05]If I'm doing anything in here, I have skills that I'm calling in the background

[00:06:21:01]that are my filmmaking skills or my editing

[00:06:25:00]skills or we were just doing color grading.

[00:06:29:10]We're going to make a color grading skill for...

[00:06:33:08]And when I say we're going to make a skill, that does not

[00:06:36:10]mean you go in and write the skill yourself. Do not do that.

[00:06:41:05]What you basically want to do is have a conversation with the agent or tell the

[00:06:47:05]agent what you want to do and then say, "Make this skill for me."

[00:06:52:12]And then you can ask it to install the

[00:06:55:05]skill and both codecs and chat GPT will do that.

[00:07:01:10]In Claude, it is under... it's under customize, I believe.

[00:07:08:20]So Claude, Cowork and chat and Claude code

[00:07:12:20]have slightly different ways they handle skills.

[00:07:15:23]But if you want to add a skill in Claude, you can hit

[00:07:19:03]add and then you can upload it or you can create with it.

[00:07:23:17]So yes, you can add skills here. So plugins, I think

[00:07:28:13]mostly over here, they just call it all plugins. I'm not sure.

[00:07:32:19]But over here in Claude, skills and plugins are a little different. A plugin is

[00:07:38:04]kind of a really robust collection of skills

[00:07:41:01]versus a skill can be kind of a small one-off thing.

[00:07:46:06]They have a bunch of skills that you can add. Like they have a bunch of pre-made

[00:07:51:17]things. But for the most part, you're going to want to

[00:07:54:23]be working with stuff that you make because that is the...

[00:08:02:09]Oh, okay. Skills. There is skills separately and

[00:08:05:10]then plugins. Okay. I swear they moved that around.

[00:08:12:01]Where was I in here? We'll just go in here.

[00:08:18:00]So when you were working with skills, like when I am doing this chat and I'm having

[00:08:24:06]it make these episodes for me with videos, every so often at the end of the chat, I

[00:08:29:16]will be like, okay, what were all the things that we learned?

[00:08:34:04]What were the obstacles we had? What do we take away from this that we can take to

[00:08:43:04]change the skill that we were using to work with this and improve it?

[00:08:47:22]So it's not just that you make these skills, you are constantly improving them and

[00:08:54:12]then they get better. But back up your original skills before you do that because

[00:08:58:14]every now and then it'll make changes where they go off the rails.

[00:09:03:20]So it's kind of an iterative process where I have a

[00:09:10:15]pretty intense set of skills by now. This is my skills folder.

[00:09:16:01]And these are all of these things that we have made. Like I have ones that help me

[00:09:22:17]with video captioning, ones that help me find tropes on Amazon.

[00:09:27:07]I have a whole protocol that every time I want to run my agents, I run them through

[00:09:37:13]a system that tells them everything that I have and

[00:09:41:22]everything that my system does and how to handle it.

[00:09:45:00]So I have ones that, just certain one off ones that my agents made me to help me

[00:09:54:18]figure out how to restore backups or some of these things I just rarely ever run.

[00:10:03:04]The agents will make a skill and be like, oh, okay, we'll just use this once.

[00:10:08:06]But there's some that I use all the time, like this creative relay skill. I had an

[00:10:12:22]intense conversation digging deeply into my

[00:10:15:15]writing process with Opus 4.6, which is a while ago.

[00:10:21:01]And we have refined this to the point where when I sit down to write and want to

[00:10:28:00]brainstorm, I load this skill and we go through this whole process.

[00:10:34:06]Like if I start a new chat here and let me get...

[00:10:40:21]Okay, so if I want to load this skill, I actually have this one installed, but I

[00:10:48:01]don't always install my skills. So a trick I learned from someone else, which is a

[00:10:53:17]good trick, is to have your skills in a Dropbox folder.

[00:11:00:04]Have your skills in a Dropbox folder and then when you want to call them, ask the

[00:11:08:08]agent, say, go to my local Dropbox, go to skills and then call the skill.

[00:11:12:22]If you do this, you can have this across several different systems and then your

[00:11:17:10]skills are completely synced across your systems

[00:11:19:23]versus when you're improving your skills all the time.

[00:11:24:08]If you have them just installed on Claude or installed on ChatGPT app and you're

[00:11:34:14]updating them on one, it might not update

[00:11:36:16]on the other.

[00:11:37:14]So, yeah, if you have your skills in a big folder in something like Dropbox or

[00:11:42:16]Google Drive or something like that and you're moving across different computers

[00:11:46:10]and you're moving across different systems or even different apps and you always

[00:11:50:10]point it to those skills, then they'll always be synced.

[00:11:55:05]So that is something that you can do. So if I want to start a writing conversation,

[00:12:02:20]I will be like, look in my local Dropbox and in my

[00:12:10:04]skills folder, load creative relay version, version two.

[00:12:19:10]That's kind of the long way around. You can still install these, in which case you

[00:12:27:03]would hit the slash command and be like creative.

[00:12:30:17]Okay, I'm not sure I actually have these installed anymore, but I do have my studio

[00:12:37:09]onboarding. If I want to start a new conversation

[00:12:39:23]and immediately have it look at everything that I am...

[00:12:44:17]A snapshot of my entire organization philosophy and how I want agents to run, I

[00:12:50:00]will go studio onboarding. And then this is calling a skill.

[00:12:56:01]So, but I don't need that here, but see the agent knows to go look in my Dropbox

[00:13:00:18]folder. That's the canonical skills folder.

[00:13:04:01]There is instructions in there that are like, these are the skills. These are...

[00:13:09:04]This is everything that we are using. So creative relay version three. This is a...

[00:13:18:08]This is a set of skills that tells the agents exactly how I want them to work with

[00:13:25:16]me when we're writing. I can call this skill in chatgpt

[00:13:30:17]and I can call it in Claude and it'll work very similarly.

[00:13:34:07]The models are slightly different, so it won't be entirely the same, but it'll be

[00:13:37:07]very similarly. So when... And because I have worked with this, I know that I will

[00:13:44:06]be like, okay, we're going to brainstorm a book.

[00:13:46:01]And then it will respond to me in a specific way because the instructions say that

[00:13:50:21]it should. So a skill will help you have a similar experience every time.

[00:13:59:04]And kind of a rule of thumb is any prompt that you find yourself putting into the

[00:14:05:09]agent again and again and again is something

[00:14:08:04]you want to turn into a skill, most likely.

[00:14:10:19]If you... If every day you go in and you're like, go into my connected email and

[00:14:16:03]find all the newsletters from yesterday and give me a summary.

[00:14:22:01]If you find yourself saying that every day, most likely you want to say it one more

[00:14:27:02]time and then say, turn this into a skill

[00:14:29:12]and then the AI will write itself a skill.

[00:14:33:13]And then either install it or put it in my skills folder. So this is... It's

[00:14:45:12]minimizing repetitive tasks. However, it's a little more powerful than that because

[00:14:50:09]I can be like load creative relay version 3

[00:14:53:11]and then load my studio protocol, which...

[00:15:01:02]Or studio onboarding. So if I have it in my studio onboarding, now all of a sudden

[00:15:05:21]it has all the access to my entire agent system. I have this whole app that's like

[00:15:14:03]a publishing command center that has everything of my business in this app.

[00:15:20:22]And when I hit that studio protocol studio onboarding skill, it immediately goes

[00:15:27:02]into this app and on boards itself with this entire agent handbook and system.

[00:15:34:05]So that's the power of this. You can take that and have it go off and do these

[00:15:40:04]different things. You can use several skills at a time.

[00:15:43:00]So now I know that it is completely onboarded in my studio and when I want to

[00:15:49:12]write, it can actually talk completely with the database

[00:15:53:09]behind my app and interact with my writing system that I have here.

[00:15:58:12]So when I'm writing, I'll be like, we're writing and it will just slide this right

[00:16:03:08]into my system here and then I can go edit it.

[00:16:05:14]So that is the kind of things that you can do with this. And these are just some

Notes

Topic times use MM:SS within this episode. Adaptable prompts are templates based on the lesson, not verbatim quotations. App names, options, and capabilities reflect the recording and may change.

Permissions have two parts — 00:00

Decide both what the agent can access and how its actions get approved.

Permissions have two parts — 00:00
ControlWhat it changes
Manual approvalYou review approval requests yourself. This adds oversight and interruptions.
Automatic approval / “approve for me”The system handles some approval decisions and may escalate actions it flags. This is different from you reviewing every action.
Skip approvals / bypass permissionsRemoves approval interruptions and gives the agent greater autonomy.
Read-only accessLimits work to inspection rather than file changes.
Workspace accessRestricts the agent's work to the permitted project area.
Full accessGives the agent wider access beyond a single working folder.

Labels differ among apps. Choose the approval setting and access scope separately; a convenient approval mode does not mean the agent needs your entire computer.

Limit exposure and keep backups — 01:51

Prompt injection is a risk in which malicious instructions inside material an agent encounters can steer it toward an unintended action, including exposing information.

A useful starting setup is a single project folder containing the materials needed for the task. Back up important files before letting an agent modify them. Wider file access makes recovery planning more important because an incorrect action can affect more than the current document.

Do not treat the instructor's personal permission settings as a requirement for using agents. A lot of useful work can be done within one folder.

Skills are reusable instructions — 05:28

A skill is an instruction file or set of files that tells an agent how to carry out a recurring kind of work. The agent does the work; the skill supplies a repeatable process.

Good candidates include:

  • Captioning videos in a consistent style.
  • Brainstorming through a particular writing process.
  • Summarizing a recurring set of emails.
  • Preparing marketing assets with the same requirements each time.
  • Loading the information and working rules needed at the beginning of a project.

A plugin can bundle capabilities such as skills and connections. Apps organize skills, plugins, and connectors differently, so the name of a menu is less important than what the installed item provides.

Turn a successful workflow into a skill — 06:33 and 08:18

  1. Work through the task with the agent and explain what you want.
  2. Correct the result until the process produces something useful.
  3. Ask the agent to turn that process into a skill.
  4. Have it save the skill in your chosen folder or help install it.
  5. After future runs, review what worked, what failed, and what should change.
  6. Keep the previous version before revising the skill so you can return to it if a change makes results worse.

Adaptable prompt:

Turn the process we just used into a reusable skill. Include the instructions that mattered, the obstacles we encountered, and how we resolved them. Save the previous version before making future revisions.

Keep one shared skills library — 10:40

Store your working skills in a clearly identified local folder, such as a synced Dropbox folder, and tell each agent where to load them. This helps avoid improving a skill in one app while continuing to use an older installed copy elsewhere.

An installed copy and a file in your shared library are not automatically the same version. Be explicit about which is authoritative and how another app should obtain the updated instructions.

Adaptable prompt:

Go to my local skills folder, load [skill name], and use it for this task.

Combine skills with project context — 12:37

An onboarding skill can explain where project information lives, how the workspace is organized, and how the agent should work with you. A task skill then supplies the specific process—for example, brainstorming or creating marketing hooks.

A repeated prompt is a useful signal: if you keep explaining the same process, it may belong in a skill. Skills save repeated instructions; your context library supplies the books, images, examples, and other material those instructions operate on.

PART 321:52

Connectors and practical author workflows

Transcript

Automatically generated transcript; may contain recognition errors. Timestamps use HH:MM:SS:FF (25 frames per second).

[00:00:00:00]examples of systems. So there's also connectors.

[00:00:05:17]And I think on Claude, I believe they're called

[00:00:07:14]connectors. ChetGPT just calls them all plugins.

[00:00:13:05]So I heavily sell on Shopify. I have a Shopify connector on here.

[00:00:22:09]I think that's in the plugins thing. So there's a lot of connectors that you can

[00:00:27:13]just get. Like a lot of these big companies

[00:00:30:18]have made connectors like this is GitHub for code.

[00:00:36:06]I have a Klaviyo connector. I don't actually have

[00:00:39:00]Klaviyo anymore, so I need to uninstall that one.

[00:00:42:17]Supabase, I use Supabase as a database a lot.

[00:00:48:19]And Shopify. So you can add like Gmail, you can add your Dropbox, Descript.

[00:00:55:01]I do a lot with Descript with Claude, where I will upload a video and I have a

[00:01:00:09]skill that it makes my captions in a specific way every single time.

[00:01:06:18]And I'll be like, load the skill, go with the video, go.

[00:01:11:21]So you can add all of these things. And when you add

[00:01:16:00]them, it gives the agent the ability to work directly with it.

[00:01:20:11]If you connect Notion, you can have your agent just basically rip through your

[00:01:25:17]entire Notion database that you give it to. You can tell it certain spaces that you

[00:01:32:19]don't want it and spaces that you do want it.

[00:01:37:21]But you can have it go through your Outlook. Higgsfield. This one is like a lot of

[00:01:43:22]the Higgsfield, OpenART, Runway. They have these connectors now.

[00:01:50:08]These are technically MCPs. MCPs...

[00:01:58:22]Coming in now, this is way easier than it was like a year

[00:02:02:01]ago, where you had to actually install these things manually.

[00:02:07:01]But an MCP is a way that your agent can talk to

[00:02:12:13]another system and call tools from it to do specific things.

[00:02:19:10]Like if you are connecting to Higgsfield, you can ask it to make images. You can

[00:02:26:04]ask it to do videos. You can ask it to do specific Higgsfield workflows that they

[00:02:31:16]have made available for the agents to use with this connector.

[00:02:36:19]These are all basically MCPs. Now they're sanctioned MCPs and plugins.

[00:02:45:02]Is the term "agents" another word for using skills? If not, how do they differ?

[00:02:50:24]The agent will use the skills, but the skills are not the agent.

[00:02:55:19]The model that is using things is the agents, but

[00:03:01:21]the agents will use skills when they're doing things.

[00:03:07:13]The skills are like the agent instructions.

[00:03:10:14]Skills are just basically small to robust instruction

[00:03:15:11]manuals for the agents to use as they are doing the things.

[00:03:19:18]There's a lot of little finicky terms here that can get a little confusing.

[00:03:29:22]The MCPs are basically a way for the system to

[00:03:38:14]connect to your agent to another company's system.

[00:03:45:17]When I'm using the Shopify, pull up my three best products this month on Shopify.

[00:04:00:04]I am using Wispr Flow because I try to dictate a lot.

[00:04:05:24]I have Wispr Flow installed on my thing.

[00:04:11:22]You can also dictate down here. I tend to like to work hands-free just because it

[00:04:18:08]saves my wrists and it's way faster and it's easier to just talk to it.

[00:04:23:05]Both the ChatGPT app and Claude have dictation

[00:04:27:18]now, I believe. I'm pretty sure Claude has dictation.

[00:04:32:08]Let me go into a new chat here. Yes, Claude has dictation.

[00:04:41:02]I had it go and pull up my three best products on Shopify.

[00:04:50:05]It's using this connector and it will tell me exactly what those products are.

[00:04:55:12]I can have it dive deep into the analytics and be like, "Okay, what are these

[00:04:59:06]products this month compared to last month or this

[00:05:02:04]month compared to last year or any of this stuff?"

[00:05:07:10]I didn't have to go take screenshots or download CSVs or any of that.

[00:05:16:00]I didn't have to do any of that because this is connected directly to my Shopify so

[00:05:20:09]it can just immediately go pull that and then give me information.

[00:05:25:00]There is a Facebook Ads MCP now. It's a little funky. Some accounts have it and

[00:05:35:16]some don't, but there's ways to connect and that's kind of complex.

[00:05:40:12]I'm not going to go totally into that, but you can connect your Facebook Ads to

[00:05:45:15]here and be like, "Okay, here's this. Here's this."

[00:05:48:14]Go take over my browser and look at my KDP data, pull that all together, and then

[00:05:54:12]give me a snapshot of the week for what my ads are

[00:05:59:03]doing, what my Shopify is doing, what my KDP is doing.

[00:06:02:06]If you're a wide author, like what your other

[00:06:05:20]dashboards are doing, things like Publishing Champ has an API.

[00:06:14:05]So API is different than an MCP, but an MCP is basically

[00:06:21:03]kind of like an API or works similarly in the same way.

[00:06:25:22]But if you hear people talking about API, that's kind of like you have to set up a

[00:06:31:22]direct connection with something. You have to get a key.

[00:06:36:04]And it's a way for your agent, again, to talk to the system, but an MCP is where on

[00:06:42:08]their end, they've already kind of configured how it's going to work versus with an

[00:06:46:18]API, you're kind of building it on your own.

[00:06:49:06]So that's kind of complex. I would stick to

[00:06:52:12]connectors for now if you're not familiar with that.

[00:06:56:03]But the connectors, they've made this way, way, way easier now. So you can just

[00:07:00:09]basically hook any of these things, and there's a lot of them.

[00:07:05:00]You can even hook it to like your Spotify and it knows all your music and

[00:07:08:16]everything. You can hook it to like fitness stuff, you can hook it to travel stuff.

[00:07:14:08]This isn't just business. There's a lot of other things you can do with this.

[00:07:21:06]So there's a lot that you can hook to. You can search in here. If whatever you are

[00:07:27:03]looking for does not have a plugin here, you

[00:07:30:13]will go and look up on Google if it has an MCP.

[00:07:35:23]And if it has an MCP, copy it and then ask your agent to help you install it. Like,

[00:07:43:03]Codex can install its own MCPs. I'm pretty sure Claude can at this point too.

[00:07:49:01]I believe it is in...it might be in the plugins area again.

[00:07:57:23]Okay, so you can create an MCP app in here, and that would be where you would

[00:08:03:04]follow the instructions for that, I think.

[00:08:06:16]And the reason I don't know where that is is because we really don't have to do

[00:08:12:11]that anymore. You can pretty much just give it to the

[00:08:15:11]agent and be like, "Do this. Help me figure it out."

[00:08:19:02]And that's another thing with working with agents is if you don't know something,

[00:08:26:01]have the agent walk you through it. And not just walk you through it. Have the

[00:08:29:23]agent help you configure or build what you need to do what you need to do.

[00:08:35:13]So that's where it's like if you're making a skill, be like, "Install it for me."

[00:08:41:06]And in Claude, it'll come up with this little thing

[00:08:43:15]and it'll have an install button, you just hit it.

[00:08:46:02]In Codex, it will just install it for you. So there's a lot of things that we just

[00:08:53:23]really don't...there's steps that we don't have to actually take anymore.

[00:09:00:02]So, and the connectors in here are, I believe, under customized. Yeah, so in

[00:09:06:20]Claude, they have skills, connectors, and

[00:09:09:22]plugins. They kind of work a little differently.

[00:09:12:13]But similarly to the ChatGPT, the

[00:09:14:11]connectors are basically like the big business things.

[00:09:17:24]And that's...it's the same type of thing. So once you add this thing, it will show

[00:09:25:01]up in your list of connectors. And if it's on, it'll be on automatically.

[00:09:32:02]Plugins...Claude does...Cowork...so Claude's a little confusing in that Cowork has

[00:09:39:22]different plugins than like if you're in Claude Code, it handles that differently.

[00:09:46:05]Or maybe not. Did they harmonize that? It used to be that the plugins were slightly

[00:09:52:05]different for each thing. But they might have

[00:09:55:01]actually harmonized that, which would be good.

[00:09:57:20]So plugins are...pluggins for Claude are kind of like really, really, really robust

[00:10:04:06]skills. And some of these are big businesses too.

[00:10:10:24]It's...yeah, it's slightly different.

[00:10:15:02]Whereas ChatGPT kind of throws them all together.

[00:10:19:18]But like you can have the Canva plugin and I believe the Canva connector too. So

[00:10:25:18]I'm not actually sure what the difference is between these two.

[00:10:30:05]But one or the other of them, you can have it hooked to Canva and then have Claude

[00:10:34:24]make Canva things for you. However, they're not

[00:10:38:07]usually very good. So you don't necessarily want to do that.

[00:10:41:12]But same sort of thing, you can just you can connect all of these things. And

[00:10:47:04]if...if...if a company has an MCP but they do not have a specific official MCP,

[00:10:58:04]then here's where you would add a custom connector.

[00:11:01:17]So, yeah, they're mostly just calling it connectors now, I believe. But same thing

[00:11:07:11]here. I have like my Shopify connected, I have Supabase. So Supabase, if you were

[00:11:15:18]working with a lot of information, and if you were a writer, you probably are.

[00:11:21:04]It's not a bad idea to have a database somewhere and Supabase is a big, big

[00:11:27:23]database site where they have a free tier and you

[00:11:32:06]can get quite a bit out of it for the free tier.

[00:11:36:03]But what it is, is it's a place to put your information where your agents can

[00:11:44:04]easily access it. Like my publishing command center here is basically a front end

[00:11:50:20]for a database where I have...I track where all of my books are and...

[00:12:01:03]Yeah. I track everything that has come out in every edition and it's all kind of

[00:12:08:02]hooked together. So I have...it's got to load. Each one of these is information on

[00:12:16:06]my books. So it'll have like the overview and the editions and the files and hooks.

[00:12:21:12]This is intense. This I built a front end because I like things that are pretty.

[00:12:26:04]You don't have to have things that are pretty. What you

[00:12:28:24]can do is have your agent go through all of your books and...

[00:12:36:07]Go find it across the internet. Go find it on Amazon. Go find it on Audible. Go

[00:12:40:11]find it in everywhere that you do your books and then have it marked down. Is it in

[00:12:45:01]hardcover? Is it in paperback? Is it in French?

[00:12:48:03]German? What language is it in? Is it done yet?

[00:12:53:01]And things like that. You can have it pull in all your blurbs, all your marketing.

[00:12:58:05]Your agent will make tables of that for you. You don't have to have a front end

[00:13:02:20]like this to be able to use that. So if you are making a database like that and

[00:13:09:22]then say you want to do marketing on Instagram and your

[00:13:15:01]database has...already has all of your hooks and all of your...

[00:13:22:05]The text of all of your books, it has...it has your release schedule. It has all of

[00:13:28:18]this. If you've put all of that in your database, when you want to go and make new

[00:13:34:13]videos for Instagram, you can be like, "Oh, go look at my database. Go read this

[00:13:41:19]book. Go look at how we've done hooks before and then come up with 20 new ones."

[00:13:47:23]And then it will do that for you. And even just me saying that, "Go do this. Go do

[00:13:53:17]that. Go do this." You can make that a skill so that you're like, you know, load

[00:13:59:13]Instagram hooks skill and then it will load that. Go look at all those things.

[00:14:06:14]And, you know, go load Instagram hooks skill with this book and it will go do that.

[00:14:12:09]Read those books, look at all those things and come back with

[00:14:15:00]20 Instagram things for you. That's kind of how you do this.

[00:14:21:01]So all of this is working toward letting your agents do things for you that you

[00:14:29:12]don't have to do yourself anymore so that you can make the high level decisions and

[00:14:37:02]not have to do the grunt work, basically.

[00:14:40:03]So like when I'm doing like that Shopify chat that we just did, and I have no idea

[00:14:47:01]where that was. Yeah, it's somewhere. But that Shopify chat that we just did, like

[00:14:55:05]I used to go through and do my Shopify, like if I wanted to add a new product and

[00:15:03:15]if I have a new book coming out, I'd need to make

[00:15:05:12]an e-book, a paperback, an audio book, a hardcover.

[00:15:08:18]I'd need to switch over all the previous books from new release to now it's not or

[00:15:17:07]put all the pre-orders, put all the little pre-order signals on everything. I used

[00:15:23:06]to do that all by hand and I don't have to do that anymore.

[00:15:26:11]So now I'm like, I'm like, you use the Shopify connector, make a new product and it

[00:15:31:23]needs to be these formats and go look at how

[00:15:36:02]I set up my other pages and do this similarly.

[00:15:39:22]And then I don't have this set up yet, but that's something I'm trying to build

[00:15:45:20]where it's like I can give it the book cover and

[00:15:47:14]it'll just go make all the mockups and all of that.

[00:15:50:22]So you can build all of this and the whole thing is setting up systems so that the

[00:15:59:14]type of tasks where you can repeat it and it can be done

[00:16:05:20]in a repetitive way that the agent can do that for you.

[00:16:09:24]So Cassie, Cassie Alexander, who runs the AI

[00:16:15:24]Marketing for Storytellers sub-stack, we run that together.

[00:16:19:20]She has all these skills set up that go to all the different vendors for ebooks, so

[00:16:27:01]like Kobo and Barnes and Noble and Google and it allows the agent to go in and take

[00:16:34:19]over her browser and upload all her books and fill out all her information for her.

[00:16:39:14]So things like that. And I have... Yeah, so you can have it like take over your

[00:16:51:06]computer for you too. Like I have been using Astra now for video editing.

[00:16:59:01]Most of this timeline was initially edited by an agent, which is kind of insane.

[00:17:05:21]But I will go into these and I have a skill set up that it knows how I like to edit

[00:17:13:22]videos now. So Astra will go over and be like, "Can I use DaVinci Resolve?" Yes.

[00:17:20:15]And then Astra will come in here and just go to town. There's actually a... If you

[00:17:27:09]do work with video, this is really good to know.

[00:17:30:22]And it's specifically DaVinci Resolve that has these abilities. You do need the

[00:17:35:24]paid version, but it's a one-time fee and then it can go to town.

[00:17:42:09]But there's scripting in here where your agent can go in and just basically take an

[00:17:49:01]entire timeline or video or whatever you're doing it with, your social media

[00:17:54:23]videos, any of that, and just rip through them.

[00:17:58:09]And then you go and look at it and you're like, "Oh, can you do it this way?" And

[00:18:02:22]it'll be like, "Okay," and then rip through it again.

[00:18:05:01]So if you do anything with video, this is a really good thing to do. And right now,

[00:18:11:02]the only model that can do that is Astra in the ChatGPT app.

[00:18:17:05]Like, if you connect your email to this or Claude, you can have it go organize

[00:18:23:23]things. You can have it summarize your newsletters or

[00:18:29:01]summarize whatever you need to do with email, basically.

[00:18:34:22]And there is... Yeah, so... You can have it organize your files on your computer,

[00:18:44:12]too, but I've had mixed results with this.

[00:18:47:04]I had Claude organize my entire download folder before, and I couldn't find

[00:18:54:04]anything thereafter because Claude's logic was not my logic.

[00:18:59:05]So if you are having it organize things on your

[00:19:02:19]computer, give it a clear roadmap and a clear plan.

[00:19:07:11]And that was several Claude versions ago, so it's very, very better than that now.

[00:19:13:08]But if you have an enormous folder with all your images for marketing for the last

[00:19:22:20]years, you can have it go in there and sort it by book.

[00:19:26:12]You can have it go in there and figure that out.

[00:19:31:04]All of these agents are able to look at video and

[00:19:35:14]images now, so they can read the text on images.

[00:19:40:01]I had a handwritten draft that I wrote, like, eight years ago in my closet, and one

[00:19:48:03]day I'm like, to hell with this being in my closet, I took my phone and I...

[00:19:53:21]It took about a half hour, but I snapped pictures of every

[00:19:56:16]single page and then had... I uploaded all of those images...

[00:20:03:10]Or put all those images into a folder and gave the folder to the Codex app here.

[00:20:12:20]I don't know if I can find that, Chad.

[00:20:17:10]Yeah, I actually think I have this chat.

[00:20:20:04]I gave everything to this and it made... it took all of these manuscript pages and

[00:20:29:17]gave me raw transcriptions and then there was a few parts where it was unclear.

[00:20:34:02]And it did this for... it was a 70,000-word book that it did in about three hours.

[00:20:42:08]It transcribed the entire book. That was

[00:20:44:07]handwritten pages. And I mean badly handwritten pages.

[00:20:48:16]So something like this. It's like I can give it that folder and I was like, "Okay,

[00:20:52:16]this is this folder. It's raw, rough draft. It's

[00:20:55:18]handwritten. Familiarize yourself with the series."

[00:20:58:09]So when I say familiarize yourself with the series, I... because I already have

[00:21:03:02]this system and everything, and I had this in a

[00:21:07:04]folder that had other parts of that series in it,

[00:21:12:20]it knew to go read the rest of that series so that it

[00:21:16:21]had the context for putting this book together, too.

[00:21:21:18]So it was able to kind of guess some of these areas that were unclear.

[00:21:29:10]So this is the kind of stuff you're doing.

[00:21:32:22]So if you have a folder with all of your series for a book in it, you can have it

[00:21:40:02]go through and be like, "Go read that and have this... here's this information."

[00:21:44:22]Or, "Go look at this and do this." Or, "It's just kind of piecing it all together."

Notes

Topic times use MM:SS within this episode. Adaptable prompts are templates based on the lesson, not verbatim quotations. App names, options, and capabilities reflect the recording and may change.

Keep the terms separate — 00:00 and 02:45

Keep the terms separate — 00:00 and 02:45
TermPractical meaning
AgentThe system carrying out the task with the model and available tools.
SkillReusable instructions for how to perform a kind of task.
ConnectorA connection that lets the agent work with another service or its data.
MCPA way to expose a system's tools to an agent, so it can call supported actions.
APIA programmatic interface to a service; a custom integration may require configuration and an access key.
PluginA packaged set of capabilities; exactly what it contains and where it appears varies by app.

A skill does not itself supply every connection it needs. For example, a repeatable sales-reporting process still needs access to the sales data.

Start with an available connector — 00:13 and 07:21

  1. Look in the app's connector or plugin directory for the service you use.
  2. Connect the relevant account and decide which workspaces or data the agent should access.
  3. Give it a small, specific task to establish that the connection works.
  4. If the service is not listed, check whether it offers an MCP connection and ask the agent to help configure it.

The class uses Shopify and Supabase extensively and also discusses email, document, design, and media services. Available actions depend on what each connection exposes. You can ask the agent to walk you through setup instead of trying to work out every configuration step alone.

Use connected data for a concrete question — 03:45

Start with a focused request, such as identifying the three best-selling products this month. Then expand the analysis:

  • Compare the same products with last month or the same month last year.
  • Combine store performance with advertising and publishing data.
  • Request a recurring weekly snapshot instead of manually collecting screenshots and CSVs each time.

Adaptable prompt:

Find my three best-performing Shopify products this month, compare them with last month, and explain the main changes.

A connector supplies access; your request determines the comparison and output you want.

Build a reusable book-information library — 11:07

A database is useful when you have many books, editions, files, and marketing assets to connect. A custom dashboard is optional: organized tables can support the work without a polished front end.

Useful information to collect includes:

  • Book and series names, overview, and source text.
  • Editions and formats: ebook, paperback, hardcover, audiobook.
  • Languages, release status, and where each edition is available.
  • Blurbs, marketing hooks, files, and release schedules.

Once the information is organized, an agent can reuse it for different tasks rather than starting from scattered materials every time.

Turn context into repeatable marketing work — 13:09

For Instagram hooks, provide the book, prior examples, and the relevant marketing information. Ask for a batch of new hooks, review them, and save the process as a skill if it becomes recurring work.

For a store release, define the required formats and show the agent existing product pages to follow. Related work can include updating preorder or new-release labels. Keep the workflow specific to the actual store structure rather than assuming every book uses the same setup.

Repeated vendor uploads are another candidate for a documented process. The goal is to offload repeated entry and formatting so you can spend more attention on the decisions and final review.

Other practical workflows — 16:39

  • Video editing: give the agent an editing process, let it assemble or revise a Resolve timeline, then review the result and provide precise corrections.
  • Email: summarize selected newsletters or organize information from a connected inbox.
  • File organization: provide a clear plan before moving files. Specify the structure you want—for example, marketing images grouped by book—because the agent's preferred organization may differ from yours.
  • Handwritten transcription: photograph the pages, collect them in a folder, and provide relevant series material as context. Review unclear passages rather than treating contextual guesses as certain readings.

Working pattern: connect the service or supply the files → explain the desired result → provide examples and context → review a first result → refine the process → turn repeated instructions into a skill.

PART 422:02

Scheduled work, creative assets, and research

Transcript

Automatically generated transcript; may contain recognition errors. Timestamps use HH:MM:SS:FF (25 frames per second).

[00:00:00:00]So I have so many different things going at any given time.

[00:00:05:02]One of the things you can have, too, is you can have scheduled tasks.

[00:00:10:14]Like, I have this specific agent go and look at what my system says needs to be

[00:00:17:01]done that day, and it does it every day at the same time.

[00:00:21:22]And here it's like it pulled the evening pass, it checked what I told it to check,

[00:00:28:18]and then it said that there's nothing we need to do.

[00:00:33:20]So I have another heartbeat here that every day it will

[00:00:43:14]check the SEO on my site, and if there's anything new, it will...

[00:00:56:14]I don't even know what this one's doing. It's doing

[00:00:58:08]something useful. I have so much going at any given time.

[00:01:01:18]I have another one somewhere. The one I was talking about was this one, actually,

[00:01:05:22]where it goes and it looks on my site, and it's just checking to see if I have any

[00:01:12:07]new products that it needs to SEO, and it doesn't.

[00:01:16:08]When I do, it will do something, but right now it

[00:01:19:20]doesn't. But it just checks every day, every day, every day.

[00:01:24:03]You can schedule these manually, but you do not have to. You can just be talking to

[00:01:30:09]the agent and be like, "Okay, this is what I want to do. I want you to check my

[00:01:34:01]calendar every day and match it against my email."

[00:01:36:10]Or, "I want you to look at my Shopify every

[00:01:39:13]day and give me a report on my analytics."

[00:01:44:11]Even to more complex tasks, I have Claude set up with a bunch of scheduled things.

[00:01:57:08]I have... I don't think it's on this one.

[00:02:00:15]It's actually in Claude code. I have a... Okay.

[00:02:06:14]Okay, in Claude code, it's called Routines. I have an agent that every day pulls my

[00:02:14:10]meta comments from Instagram and Facebook and hides any anti-AI comments and

[00:02:21:03]then... or no, it deletes any anti-AI comments and then...

[00:02:26:10]So, when I go in to actually look at my comments, I never, ever see them. So, that

[00:02:32:03]is really nice. So, you can set up things like that. That's a little more

[00:02:35:04]technical. I actually had to set up a Facebook

[00:02:38:07]app that connected with it to do all of that.

[00:02:43:06]I actually have a set up here somewhere in here. I build these intense systems and

[00:02:51:22]then sometimes I don't even know where the things

[00:02:54:13]are inside of them. So, yeah, it's somewhere in here.

[00:02:59:09]But you can do these things. Like, I needed for a

[00:03:06:19]Kickstarter to... I needed to make a map of a location in my book.

[00:03:15:23]And I'm like, "I don't want to spend all this time rereading this entire series to

[00:03:21:15]figure out where everything was in this palace in the book."

[00:03:26:09]So, I was like, "Okay, go build me an architectural app that the agent can interact

[00:03:36:14]with and read through all of my books in this series and then basically make a

[00:03:42:09]layout of where you think everything is."

[00:03:44:07]This was not the initial layout. I had to kind of tweak this. But over here, you

[00:03:49:11]can see it's showing where everything it found in the

[00:03:56:03]books was and it's reasoning for where it put things.

[00:04:00:06]So, things like this. Do I use this all the time? No, but it is useful. So, I was

[00:04:07:18]able to take this and turn it into an actual map for the book.

[00:04:13:07]Another thing is that the Codex

[00:04:22:24]app will... It automatically has the GPT image...

[00:04:29:17]2.5 now integrated with it. So, you can... I needed

[00:04:36:13]YouTube thumbnails. So, I was like, "Here is my YouTube video."

[00:04:42:13]And I think I have these character references for all my characters and I gave it

[00:04:49:24]that folder and I gave it... This was the

[00:04:53:24]thumbnail that I originally had that wasn't really working.

[00:04:56:20]And I'm like, "This is this. This is that. I want to use these characters. I want a

[00:05:02:20]winning thumbnail and give me several versions."

[00:05:07:05]So, it looked at all of that and was like, "Okay, let's try these." And then it

[00:05:13:13]gave me these new thumbnails. So, I have been AAB testing those on YouTube.

[00:05:18:09]And it still has this in context. So, I think... Yeah, I kind of went off on a

[00:05:28:22]tangent with something else. But then when I had to come

[00:05:31:00]for the next video, I'm like, "Okay, we need this one."

[00:05:35:02]And we wanted similar writing. So, then it made me the thumbnails for episode two.

[00:05:43:21]So, I wanted to make a map of my fantasy kingdom.

[00:05:48:22]I had it read all four of the books again and come up with, "Okay, what are the

[00:05:54:03]major landmarks?" It came up with this

[00:05:57:19]blobby thing and I'm like, "Well, that's a start."

[00:06:00:17]So, I refined it from there. I'm like, "Okay, this, this, this, and this. Go back

[00:06:06:01]and change it." And then it's going through

[00:06:09:23]and it's like this back and forth process.

[00:06:11:20]This was the next iteration. That's way better. Still not quite it. So, I actually

[00:06:17:20]took a screenshot of this and drew over top of it in

[00:06:21:03]Photoshop and then brought it back in in one of these areas.

[00:06:27:06]I brought it back in. This is my drawn over version. So, I'm like, "Okay, this is

[00:06:32:12]how we want to change this." So, I'm like, "Can

[00:06:37:07]you make a SVG of that?" And it came back with this.

[00:06:41:14]And I'm like, "Wow, that is actually really good." So, I think we refined it a

[00:06:47:21]little bit more. I'm not even sure. We might have just ran with it then.

[00:06:51:00]So, this is just ChatGPT coding this, or Astra coding this to make it look like

[00:06:59:12]that, but I wanted it pretty. So, I'm like, "Use this as the reference to, I think

[00:07:05:24]this is the one we actually ran with, to make this map here."

[00:07:09:11]So, we started from my books and ended up with this. So, this is the kind of stuff.

[00:07:15:13]Like, you don't have to reread your books every time you do it, or you don't have

[00:07:19:05]to go search in through your email. You

[00:07:20:21]don't have to be looking all of these things up.

[00:07:25:03]You can just... The power of this is that it

[00:07:28:13]can rip through these things and then just do it.

[00:07:30:07]So, yesterday, I've been running this promotion thing where... Running this

[00:07:36:07]promotion thing on my website where I have a Harvest Festival in one of my books.

[00:07:41:21]So, I'm running this thing on my website where everyone's filling up seats in the

[00:07:45:22]Great Hall and there's milestones. So, it's

[00:07:48:21]kind of running almost like a Kickstarter.

[00:07:50:06]Like, 25 seats that are filled up get this thing for everybody and 50 gets this

[00:07:56:02]thing. Well, we just passed 50 and I'm like, "Okay, I need a goodie for everybody."

[00:07:59:18]I have not planned this yet because I've got so much going on. Oh no! What do we

[00:08:04:02]do? And I'm like, "Okay, we're gonna make tarot cards."

[00:08:06:01]So, I did not spend a ton of time in Mid-Journey making tarot cards. What I did was

[00:08:12:16]come in here to this chat that had already read through all of my books.

[00:08:16:24]And I'm like, "Okay, I wanna make tarot cards in the main

[00:08:19:00]characters. These are the main characters in my series."

[00:08:22:16]This is what I want you to do, proposed designs.

[00:08:26:20]So, Astra, who had already read through all of my

[00:08:30:15]books, was like, "Okay, this is what I think we should do."

[00:08:34:21]And then, we back and forth a little bit because I didn't

[00:08:38:17]like one thing. And then, these were the first versions.

[00:08:43:03]And I was like, "It's okay. They're not really

[00:08:47:10]looking like what I vision the characters to look like."

[00:08:50:03]But, I happen to have this folder with all of

[00:08:54:07]these references I've already made for my characters.

[00:08:57:24]So, let's go back and I picked those all

[00:09:02:19]out, dumped them in here, and re-rolled them.

[00:09:05:11]And so, it re-rolled to the likenesses that were exactly on those images. And I'm

[00:09:10:14]like, "Yes! That is amazing! We're gonna roll with it!"

[00:09:13:10]So, I made a folder of it. I think I just redid this

[00:09:17:16]one. That was it. Because it needed to add magic to it.

[00:09:22:16]But, we made a folder of it, uploaded it, and boom! Sent it

[00:09:26:15]out to everybody. And I think that took about half an hour.

[00:09:30:07]So, if you build...

[00:09:35:23]The thing with working with agents is systems

[00:09:39:11]and context and knowing how to ask questions.

[00:09:45:11]If you have the skills, like things like the skills are your system.

[00:09:51:05]So, it's like if you have repeatable systems that will

[00:09:53:15]help you get faster and help you have more accurate work.

[00:09:56:24]Take away your brain space cycles because you don't have to think about...

[00:10:00:15]You don't even have to think about how to ask the right question.

[00:10:05:09]The question is already asked for you in your directions. You just need to know

[00:10:11:15]which directions to call to ask the question.

[00:10:14:24]So, if you have the systems, you have the context, like things like these...

[00:10:22:04]This is context. So, if you have images of your characters, if you have your books

[00:10:30:17]in a folder, if you have documents of marketing, if

[00:10:35:17]you have research that you have sent your AI out to do.

[00:10:40:06]It can take all of that and synthesize it into new

[00:10:43:06]things that you need. Like, I needed these tarot cards.

[00:10:46:20]I had all of the context. I had the books as context. I had the images as context.

[00:10:54:21]I didn't have a skill for this one specifically, but

[00:10:57:02]ChatGPT has its own image generation skills that it works through.

[00:11:03:02]So, this was the context that I was able to

[00:11:05:01]make this very quickly. And they look great.

[00:11:09:02]So, it had all of that together. And then

[00:11:13:24]the ability to ask the right questions is...

[00:11:17:24]That's something that you will get as you work with agents more.

[00:11:23:05]Where, like, when I am starting a new chat here, I'm not starting this as...

[00:11:30:20]This is my skill for the map. I sent it to do this studio onboarding so that it

[00:11:36:23]immediately has access to every single book in my system.

[00:11:41:18]It has access to every single book in my system. I have all of this context in

[00:11:50:20]here. I have all of my Kickstarter voice, marketing voice, newsletter voice.

[00:11:57:09]I have had my agents rip through all of my

[00:12:00:23]series and come up with algorithms on how I plot.

[00:12:05:16]I've had it go through and grab all my style. I have all of

[00:12:10:16]this. This is, like, years worth of context I have built up.

[00:12:14:13]I have all this story bible context for every single series.

[00:12:18:23]And, again, this is the whole thing that I've built to kind of interact with this.

[00:12:25:04]But it's not loading here.

[00:12:29:06]This is something that you can build over time. You can make a database of this

[00:12:34:18]stuff. You can have it in folders. You can

[00:12:36:14]have it in Notion. You can have it in Obsidian.

[00:12:39:15]This is the way I've chosen to organize my context, which is basically all I'm

[00:12:44:16]doing. This is one big organization for all of my context that when I want to send

[00:12:51:14]my agent to do things, I can be like, go look at this book

[00:12:54:23]and the image context in my super base and pull that together.

[00:13:01:17]I have all my world building data here. I have all my

[00:13:06:04]costumes and objects. This is kind of not all the way done.

[00:13:13:05]And they're loading kind of slowly.

[00:13:15:21]But I have marketing context. I have, like, the tropes and quotes. And I had the AI

[00:13:25:02]go through everything that I have and pull out all the quotes, all the quotes in

[00:13:33:12]the books so that when I am making Instagram slides,

[00:13:40:03]I can go through and be like, oh, we need this one.

[00:13:42:21]Oh, we need this one. Oh, we need this one. So that kind of thing. And also, one of

[00:13:52:05]the things agents are very, very good at is pinpointed research. Fable in

[00:13:58:11]particular in Claude. If you are in Claude and you go back to Cowork.

[00:14:06:11]If you were in Claude Cowork and you go to Fable and you are like, go look at or

[00:14:16:04]go research all of the problems that readers have with this genre and look at

[00:14:26:20]Reddit and look at Goodreads and look at this and look at that.

[00:14:29:21]And you'll find like reader pain points. They'll be like, oh, we don't like these

[00:14:33:21]endings or oh, we want more of this or oh,

[00:14:36:09]there's no books that have this trope in it.

[00:14:39:03]You can find that out very quickly and then be like, OK, here's a gap I can feel.

[00:14:45:06]Here's a gap I can fill. But that's just like that. You can do like multifaceted

[00:14:51:19]research where, say you do that and you get this big report.

[00:14:55:22]Then you have your A.I. put that in your database. Like I will send a A.I. out. I

[00:15:02:16]sent the A.I. out after micro market research. And this was one big run on like go

[00:15:08:19]go look at Goodreads, go look at Amazon, go look at Reddit.

[00:15:11:13]Find all of these trending micro markets that that people are talking about, but

[00:15:19:23]they're not like big. They're the niche things. So

[00:15:22:07]it came back. This entire thing was one research run.

[00:15:26:09]It came back on a massive trend report and then a deep dive into every single one

[00:15:30:23]of these niches. This is one research call.

[00:15:35:12]If you tell it if you tell your agent to especially in Claude, if you tell it to

[00:15:42:18]start a workflow, it'll spin up like 100 agents or something like that.

[00:15:49:24]If you do that, know that it's going to really crunch your usage, but it's going to

[00:15:53:24]come out with just craziness. So I dump this in my database. So the next time I

[00:16:00:19]want to do a massive report, I'll be like, go look at it.

[00:16:04:20]Go look at this report in my database. Now go look at YouTube, go look at Amazon,

[00:16:11:10]go look at all of these other things like this. This one was a brand research run

[00:16:17:23]that I had it do. I had it look at I had it look at my micro market report.

[00:16:23:04]I had it look at I pulled in the entirety of my Facebook ad library into here.

[00:16:30:06]Every ad that I have ever run is now in this database. Every newsletter I've ever

[00:16:35:07]sent is in this database. Go look at this, go look at that, go look at my

[00:16:39:09]analytics, go look at this, like the ads that have done the best.

[00:16:46:15]Now go look at YouTube and here's three YouTube videos I like. Go pull the

[00:16:51:01]transcripts. So you're able to synthesize these large amounts of information. So I

[00:16:58:12]have all this information here. Now I want to go make my ads.

[00:17:02:01]I'm like, go look at my brand report, go look at my ads, go look at my Shopify

[00:17:10:14]analytics, go look at my current ads that are performing now, go look at my

[00:17:16:03]Instagram, go pull from this, go pull from that.

[00:17:19:08]Now we're going to synthesize these ads and when I'm doing the ads, I have skills

[00:17:24:21]that I have skills that are putting together the ads the way I want them to be put

[00:17:31:11]together like that's my brand voice. This is my brand voice.

[00:17:35:11]So this is what I mean when we're building systems that can pull all of these

[00:17:41:24]things together because it's not about doing it is

[00:17:44:15]about doing things faster. It is about doing things faster.

[00:17:48:01]It's also about doing things very smartly because these kind of reports are the

[00:17:54:07]kind of reports that agencies spend thousands of dollars hiring people to research

[00:17:59:17]to do. And here you can just do it in a couple hours in an

[00:18:04:05]afternoon and have this information ready to hand to use.

[00:18:10:01]I have not read all of this. I've skimmed over parts of it, but I have not read all

[00:18:14:18]of this, like this micro market report. I looked at areas that were interesting to

[00:18:20:23]me, but I didn't read all of this. However, when I am choosing what genre I want to

[00:18:29:13]write in, when I'm planning future books, I'm like, go look at this report.

[00:18:35:06]Go look at all of these. And because my system has all of my context in it, like

[00:18:45:00]what I call like my author DNA, my system knows what I like. So it has the context

[00:18:53:18]of these are the types of books that I write. These are the types of books that I

[00:18:57:03]consistently do not write or I just like I don't kill

[00:19:00:08]my characters. I'm never going to write Game of Thrones.

[00:19:02:24]So that kind of thing, it's pulling all of this together. It is synthesizing mass

[00:19:10:05]amounts of information. So when I'm doing my Instagram, when I'm doing my ads, when

[00:19:16:09]I'm making businesses decisions, I can make

[00:19:18:17]them with a high degree of information behind them.

[00:19:24:24]So it can help you do smaller tasks. It can help you do like this in scope is kind

[00:19:34:05]of a smaller task. But if I had to like sit down and draw this on my own, for one,

[00:19:41:16]I procrastinated doing this for nine

[00:19:43:08]months anyway, it just would not have been done.

[00:19:45:08]And then when I finally did do it, the agents were smart enough to be able to rip

[00:19:51:12]through and get all of this for me. So this is the map that I wanted. But I didn't

[00:20:01:14]have to I didn't have to exhaust myself getting there. And I could have drawn this

[00:20:08:10]by hand, but this is going to be a better result.

[00:20:12:12]So these are the kinds of things that we can do. And there's there's just so much.

[00:20:20:21]So anything you don't know, you can send it out to research to find out any

[00:20:26:03]anything that you want to build, you can have it walk you through and or build it

[00:20:32:01]for you and walk you through how to set it up.

[00:20:35:05]A bunch of things that you don't want to do on your computer, like you can just

[00:20:40:24]have it like like like video editing. It's not that I don't want to do this. It's

[00:20:45:20]just I'm always trying to find ways to make

[00:20:48:19]this easier. I had I think it was over 200 clips.

[00:20:54:00]There's a there's a lot of clips in here. And I'm like, we need to get these all on

[00:20:59:06]the timeline. And and the agent was able to look at the systems I'd already built

[00:21:04:12]around where the clips were and line them all up all

[00:21:07:24]in order. And I went through and tweaked it by hand.

[00:21:11:07]And then I'm like, OK, like here's some spots and and we need to compress this and

[00:21:17:03]whatever an agent's like, OK, it goes in and

[00:21:19:08]just put soup, soup, soup, soup, soup around.

[00:21:21:21]It's not perfect, but it's a back and forth process.

[00:21:25:06]And that's kind of that's kind of what it is working with.

[00:21:30:04]Yeah, it's not going to be perfect. It's you can get it to to write email in your

[00:21:35:21]voice. You can get it to do all of these things very closely to how you want to do

[00:21:40:19]it. You can sometimes hit it 100, 100 percent.

[00:21:43:19]But you're not going to get 100 percent all the time, all the time on everything.

[00:21:48:19]It can get very close, but you always have to look over

[00:21:51:08]everything. But most of it is it's just going to be back and forth.

[00:21:55:22]It can be like, go do this thing and it'll do the thing.

Notes

Topic times use MM:SS within this episode. Adaptable prompts are templates based on the lesson, not verbatim quotations. App names, options, and capabilities reflect the recording and may change.

Schedule a defined process — 00:05

A scheduled task runs a specified check or workflow at a chosen time or interval. The examples use heartbeats and routines for daily checks, reporting, and maintenance.

Define the trigger and what should happen when there is nothing to do. For example:

  • Check for new store products each day.
  • If there are no new products, take no action.
  • If a product needs attention, apply the established SEO process.

Other examples include comparing a calendar with email, producing an analytics report, and moderating comments according to chosen rules. More involved integrations may require additional setup; the comment-moderation example required a connected Facebook app.

Adaptable prompt:

Every day at [time], check [source] for [condition]. If there is new work, use [process]. If nothing has changed, report that there is nothing to do.

Make creative assets from existing context — 02:59

The agent needs both the source material and a clear description of the result. Existing books, character references, prior designs, and examples make it easier to create something specific to your work.

Make creative assets from existing context — 02:59
AssetInputs to provideWhat to review
Building or location layoutRelevant books and descriptions of the settingRoom placement, relationships between locations, and the evidence behind the layout.
YouTube thumbnailsThe video, character references, the previous thumbnail, and the desired directionCharacter likeness, consistency with the video, and differences among proposed versions.
Fantasy mapBooks, landmarks, and corrections to the initial layoutGeography and placement before developing the finished visual style.
Character cards or reader rewardsStory context, intended characters, and visual referencesCharacter identity, likeness, thematic details, and the final set.

Use a revision loop for visuals — 05:43

  1. Ask for an initial interpretation based on the source material.
  2. Identify specific things that are wrong or missing.
  3. Request a revised version.
  4. When words are not enough, annotate a screenshot or draw the intended change.
  5. Use the corrected layout as the reference for the next version or final image.

The map workflow moved from book research to a rough layout, revisions, a hand-marked correction, an SVG, and a finished map. The card workflow improved after adding established character references. The transferable lesson is to provide better evidence and concrete corrections rather than repeatedly asking for a generically “better” result.

Build a context library over time — 09:35

Three useful ingredients are a repeatable process, relevant context, and a clear request.

Collect material that can serve more than one task:

  • Book manuscripts and series bibles.
  • Character, costume, object, and setting references.
  • Worldbuilding details.
  • Writing, newsletter, and marketing voice examples.
  • Tropes, quotes, blurbs, and prior campaigns.
  • Research reports, advertising history, and analytics.

Use folders, a database, Notion, Obsidian, or another system you can keep organized. The storage choice matters less than being able to tell the agent where the authoritative material is. You do not need to build a large custom application before beginning.

Ask focused research questions — 13:52

Research is more useful when it has a specific question: reader frustrations in a genre, desired endings, underserved tropes, or promising small niches.

A practical sequence is:

  1. Define the genre, audience, or question.
  2. Gather information from relevant reader and market sources.
  3. Save the resulting report in your context library.
  4. Reuse it alongside your own books, brand preferences, ads, newsletters, and performance data.
  5. Ask for a synthesis tied to a decision, such as campaign ideas or directions to explore for a future book.

Large, parallel research workflows can consume substantial usage. Match the size of the research run to the question you need answered.

Keep the decision and review with you — 18:10 and 20:35

A large context library lets you ask more informed questions, but it does not guarantee a finished result on the first attempt. Use your preferences to constrain the options: the kinds of stories you write, outcomes you avoid, visual standards, and brand voice.

For creative and editing work, expect a cycle of draft → inspect → give specific feedback → revise. An agent can assemble a substantial amount of work quickly; you still need to look over the result.

PART 520:33

Building your system, Q&A, and next steps

Transcript

Automatically generated transcript; may contain recognition errors. Timestamps use HH:MM:SS:FF (25 frames per second).

[00:00:00:00]I have had like this SEO chat. I think that's a maintenance one, but I had it audit

[00:00:08:05]my entire Shopify SEO strategy and it went through and it looked at everything.

[00:00:14:22]And it we had a bit of back and forth about how I want to run my SEO and then we I

[00:00:24:16]had it just ripped through my entire back end of my site.

[00:00:27:21]And like two hours later, it was like, OK, done hundred. Was it two hundred and

[00:00:33:08]eighty four native SEO records, one hundred eighty three live passes.

[00:00:38:11]I was able to get rid of a plug in that I had been paying for. So and that's where

[00:00:43:11]I set it up to every day. It looks for new products.

[00:00:46:01]And is there a new product? No, then we don't do anything.

[00:00:48:17]If there's a new product, then we go in and change that SEO according to these

[00:00:52:19]guidelines. So this is the kind of stuff we can do.

[00:00:57:11]It's just there isn't any one thing where it's like, oh, you have to do this or or.

[00:01:04:13]It has to be done a certain way or you need to do systems

[00:01:06:24]a certain way. Everyone's going to have different systems.

[00:01:10:24]Every way you work with this is going to be different and kind of the only way to

[00:01:16:17]figure out how it's best for you is to just kind of start.

[00:01:21:18]My systems are intense at this point because I've been building them for years. But

[00:01:30:05]I mean, even even before there were agents,

[00:01:31:22]I had folders full of all of these things.

[00:01:34:19]So. It's just a matter of you just got to start. So like a good place to start with

[00:01:41:22]this, like for for help with your books would be to make folders of your books.

[00:01:47:06]Like if you have like wherever you format things, make a copy of that folder and

[00:01:53:05]then have the agent go through and pull all your

[00:01:56:05]final versions of your books into into folders.

[00:01:59:06]Or if you want to be brave, get a Supabase and have it start pulling it into a

[00:02:03:06]database or notion or obsidian or something like that.

[00:02:07:08]Any place where you have a consolidated area where the A.I. can go to look for

[00:02:12:16]context and then start gathering your images

[00:02:16:01]that you've been collecting throughout the years.

[00:02:19:06]Pull that into context and have it go through your

[00:02:23:06]Instagram or go through your TikTok and get get your descriptions.

[00:02:30:08]Anywhere where you have context, have it start gathering that. Have it go have it

[00:02:35:14]go look at all your books on Amazon and pull all your descriptions.

[00:02:39:10]Agents sometimes flake a little on Amazon. So if you have wide books, I would send

[00:02:43:21]it to like Barnes and Noble or Kobo first.

[00:02:46:21]To go pull all your descriptions, pull all your blurbs. It can harvest your

[00:02:52:07]reviews. It can all of this. You gather it into context because then like say

[00:03:00:06]you're making marketing stuff and you're like, oh, I would normally have to go to

[00:03:05:19]Amazon right now and brave looking at my reviews to find a review to put on this

[00:03:09:23]quote on this thing that I want to put on social media.

[00:03:13:04]But if you've already had it go and harvest your reviews, you don't have to do

[00:03:16:21]that. You'd be like, OK, pull three of my best reviews and then put them on these

[00:03:21:04]images and then go look at my character images and then make something like it and

[00:03:26:00]then put it all together and then you've got your ads.

[00:03:29:23]So.

[00:03:32:04]And then as you go, you just build more context, you build more research, you build

[00:03:38:00]more systems so and you build more skills. So it's like as you're doing this, if

[00:03:43:06]that's something you do all the time, you're like, you've done that three times in

[00:03:46:07]this chat and you're like, OK, turn this into a skill so that the next time you

[00:03:49:17]don't have to say all of that, you'll just be like, go do this with this book done.

[00:03:53:18]So that is how to work with the desktop agents. And that's why they're so powerful.

[00:04:02:08]And and and I mean, so like even like with that Instagram making quotes of this and

[00:04:08:15]that with all that context, if you turn it into a skill and then you have it

[00:04:11:18]scheduled so that every week you get a new

[00:04:14:23]batch of things made for you automatically.

[00:04:17:02]You just happen to pop into this folder and you just go look at that folder and you

[00:04:21:12]can even set up systems where it pulls that for you

[00:04:24:09]and schedules it for you. So all of this because like.

[00:04:30:20]We're trying to do things smarter. We're trying to do things where.

[00:04:38:09]We can reclaim back the creative time that the admin admin had us doing and also.

[00:04:46:20]Quite frankly, because lots of the marketing world is doing this, it's not

[00:04:52:09]necessarily it's still an advantage because not everyone is doing it, especially in

[00:04:56:14]the author world. It's still an advantage.

[00:04:58:11]But there's going to reach a point where everyone is doing stuff like this, like,

[00:05:02:08]especially in the high power marketing realm, you are getting people who are doing

[00:05:06:10]things like this heavily where this is kind of the new the new layer.

[00:05:12:10]This this is getting to be the new

[00:05:15:01]standard and it's going to get easier because the.

[00:05:20:03]The systems that people were building are continually getting easier to work with.

[00:05:25:14]But this is kind of how you work at scale and these are you build these systems and

[00:05:31:22]then you can scale them and you're able to take

[00:05:36:21]what would take you a much smaller sphere of operation.

[00:05:44:15]I guess you're able to take that and expand it and do a whole lot more.

[00:05:48:18]You might not even be reclaiming back a whole lot of time. You can just do a whole

[00:05:53:05]lot more. You can do more with your ideas. You

[00:05:55:24]can you can do more things and do them creatively.

[00:05:59:18]You can make your special editions while you are making your audio books, while you

[00:06:03:14]are making your marketing, while you are making all of this.

[00:06:08:16]You can have these things done and it's it's just it's building systems and it's

[00:06:15:23]learning to think in systems and learning to think it's it's almost like in the sci

[00:06:22:00]fi movies when when they're manipulating holograms and

[00:06:27:10]like Iron Man with his Jarvis and that kind of stuff.

[00:06:31:02]That is that is what we are doing. Like it's it's literally sci fi what we're

[00:06:35:04]doing. Like it's those of us who work with it every day. It's like I think we just

[00:06:39:03]get into the habit of just like, oh, we're just doing this and that.

[00:06:41:14]But if you really step back, it's like, no, this is really

[00:06:44:15]intense and it's it's this it's new and it's still evolving.

[00:06:48:15]If we're just getting started, can you recommend

[00:06:50:15]your top three suggestions for us to do with an agent?

[00:06:55:16]So everyone's top three is going to be different. I think I would look at.

[00:07:02:06]I'd make a list of the top top like 10 things that you were doing that one you

[00:07:08:08]don't like to do or you think are tedious.

[00:07:13:06]And or you think take too much time. And that is where I would start. So like like

[00:07:22:05]if you don't like to clean your your email

[00:07:24:08]inbox, that would be a very good place to start.

[00:07:27:12]Hook your Gmail to it and have it have it go through and clean up your inbox and

[00:07:34:20]then have it schedule it to clean up your inbox in

[00:07:38:15]the same way like every week or something like that.

[00:07:43:10]Which scheduling that might be a little like don't don't don't run it autonomously

[00:07:49:01]until you know that it knows what you want.

[00:07:51:11]Otherwise, it can just go through and be like, oh, I'll

[00:07:53:10]just delete all these things for you. Yay. And it's like, no.

[00:07:57:23]But so like something like that newsletters

[00:08:02:12]newsletters is a really good place to start.

[00:08:04:12]So like I have. Oh, I think it's over here. I have I have a connector to Spoks,

[00:08:13:21]which is what I do my newsletters through.

[00:08:16:07]So every time I want to do a newsletter, I'm just like, OK, like go pull my

[00:08:22:07]products from Shopify and figure out this and that and then write me a newsletter.

[00:08:29:17]And I already have I had it go through all of my previous newsletters at some point

[00:08:36:17]and pull out like what my newsletter voices.

[00:08:39:09]I had it make a newsletter skill so that it

[00:08:42:06]writes in my voice and it knows what I like.

[00:08:45:06]It knows the type of things that I do. So when. So like, OK.

[00:08:53:15]I think this will open the box link. It'll it'll just because it's connected to

[00:08:58:18]Spoks, it'll write that newsletter and then it'll just immediately put it in here

[00:09:02:18]so that all I have to do is proof it and then be like, go.

[00:09:07:16]I think this is when I sent already, so

[00:09:09:02]I'm not sure it's going to let me. OK, yeah.

[00:09:11:17]So I had it do that and it automatically pulled the

[00:09:15:18]product and the links, wrote the wrote the what do you call it?

[00:09:21:22]The accessibility text. It made the buttons. It wrote the copy. It pulled this in.

[00:09:28:18]It's like I gave it a brief on what I wanted, but then it pulled these things in

[00:09:33:07]for me and even went and found this story that I wanted to add in here.

[00:09:37:03]So all I was able all I had to do was proof this and then hit go.

[00:09:43:21]So considering an agent for going into a six to claim

[00:09:47:04]all the books I just translated, that is a very good idea.

[00:09:50:03]Actually, if you have 100 books to claim, would it be

[00:09:54:01]safe for the agent to live in your account to do the things?

[00:09:58:00]Generally, yes. However, some.

[00:10:03:00]I think you get more into trouble with it's not even that you get into trouble.

[00:10:07:13]Some sites don't like agents. Amazon sites can be kind of

[00:10:12:05]notorious for bouncing agents where it's nothing bad happens.

[00:10:18:02]It's just that the agent won't do what you want it to do.

[00:10:21:11]The agent will be like, I can't do this thing. Help. Oh, no.

[00:10:27:01]It's iffy like like for the for the Amazon sites. It's iffy.

[00:10:34:18]I haven't let it into my KDP back in yet

[00:10:37:20]just because I'm slightly paranoid about that.

[00:10:40:13]But like I will have it go on the front end of Amazon and rip through things.

[00:10:46:18]If the agent what I would do with that is you can like in the connector.

[00:10:53:17]You can record a skill where you record it. You do it by hand.

[00:11:00:07]It watches what you're doing and then it knows the steps so that it can go through

[00:11:07:00]and repeat those steps when you're doing it.

[00:11:11:11]And you don't necessarily have to do it like that either.

[00:11:14:03]You can kind of talk it through as well.

[00:11:17:21]But. Yeah, it's.

[00:11:23:10]It's one of those things where.

[00:11:27:16]You just kind of have to walk through it and figure out.

[00:11:34:03]If it can do it for one and make sure that it does it the

[00:11:37:19]right way, you kind of have to babysit it the first few times.

[00:11:40:16]If it gets like two or three books and you're good and you're golden and you watch

[00:11:44:14]it work and and then you're like, OK, now I

[00:11:47:22]can go and let it do like the next 50 of them.

[00:11:52:00]And just check occasionally. That's kind of that's kind of what you have to do with

[00:11:56:06]any kind of agent automation where it's like the first time the first time you have

[00:12:00:11]to show it what to do the next few times.

[00:12:03:05]You need to babysit it and occasionally correct it.

[00:12:05:23]You're like, no, that's not what we're doing. We need to go do it this way.

[00:12:10:02]And then once you have seen that it can repeat the process successfully several

[00:12:17:14]times, then you can usually let it go off on it.

[00:12:21:12]So but still check on it.

[00:12:24:22]The agents these days are way better than the agents of a few months ago.

[00:12:31:03]If you were doing anything with a high degree of complexity, use Astra or I'd use

[00:12:37:15]Astra or maybe Sol as the models, these models here, possibly Fable.

[00:12:47:02]With Claude agents Fable and Opus, I wouldn't hand Fable can handle really, really

[00:12:53:04]intense tasks like that, but Fable sometimes refuses to

[00:12:56:15]do things where it thinks it's it's hitting safety filters.

[00:13:01:19]So watch that. But Opus is not a great agent for like it can do these things, but

[00:13:09:05]Opus kind of spins on its own wheels sometimes.

[00:13:11:23]So I would not hand this kind of task to Opus in Claude.

[00:13:17:22]So since Goodreads is owned by Amazons, can you use an agent to clean up a

[00:13:21:19]Goodreads that has old outdated information and

[00:13:24:04]book versions or is that something you can't do?

[00:13:28:06]That you could try. I think I've heard people having more luck with Goodreads than

[00:13:36:11]Amazon, but it just it depends. It just depends. You can try.

[00:13:42:20]Everything is worth a try. It's totally worth a try.

[00:13:46:22]And and and seeing if the agent is able to manipulate that on its own.

[00:13:52:24]A lot of most sites will let agents in at some point, but and when an agent is like

[00:14:04:04]working on things, you'll see the browser

[00:14:05:20]moving around and you'll see like the cursor moving.

[00:14:09:14]I would try with Astra. See if Astra can do it.

[00:14:12:23]If Astra is like, oh, their site is blocking me.

[00:14:17:09]If it doesn't work with the browser, have it take over

[00:14:20:06]your computer and and it might be able to do it that way.

[00:14:23:20]I've every now and then if something doesn't work with the browser, I have it.

[00:14:28:01]It must approach it slightly differently because I

[00:14:31:13]sometimes have more luck when it takes over the computer.

[00:14:36:04]But it's worth a try. It's.

[00:14:40:02]I know like Cassie has had it uploading to her KDP dashboard, so it's like it's

[00:14:47:05]it's possible and and most of it like if you hit

[00:14:51:10]walls, be like, OK, like how do you approach it differently?

[00:14:54:21]Can can you do it differently? Because a lot of times the agents are like, oh,

[00:14:58:00]we'll just scrape this whole thing with Jason.

[00:14:59:20]It's like, nope, you need to act like you are actually a person clicking on things.

[00:15:04:15]So sometimes it'll go off and try to do things and it'll bounce and you'll be like,

[00:15:09:19]nope, go go do this a different way or nope, go do this this way.

[00:15:14:19]So it's worth a try. I think you might be able to.

[00:15:20:01]It it's and if you do that with like old outdated information, you would first need

[00:15:26:15]to have you need to give it the right information or you'd need to tell it where to

[00:15:31:01]look to where it's like, OK, go look at my Amazon.

[00:15:34:23]And this is the current information. Update it to this or go look at my my

[00:15:40:16]database, which has all this stuff in it and then update it to this.

[00:15:43:24]So. You kind of have to give guardrails. You need to give instructions.

[00:15:50:16]Otherwise, it'll kind of just go off on its own and pull from any random

[00:15:54:00]information that it thinks is helpful, which is not always helpful.

[00:15:57:14]Like the agents right now are very, very,

[00:15:59:09]very smart. However, they still make mistakes.

[00:16:03:05]So and if you if you do have it do an automation like that, just double check

[00:16:08:00]everything, especially if it's public facing.

[00:16:10:03]So, yeah, I I think I would give that to Astra and and and see I haven't personally

[00:16:20:14]had it going into Goodreads because I am allergic to Goodreads.

[00:16:25:03]I just don't go there. But I think that's worth a try.

[00:16:30:14]Um. Which is is is any of it, it's like.

[00:16:38:02]Things things can change often and sites can change often and things that you could

[00:16:44:12]do a couple of weeks ago, sometimes you

[00:16:47:00]can't do now, it's it's all kind of in flux, so.

[00:16:52:03]So it's one of those things where you just kind of remain fluid with things.

[00:16:58:06]But you yeah, you just have a conversation as like, can we do this? Can we do that?

[00:17:04:18]And how do we do it? And and if you run up

[00:17:08:13]into obstacles, how do you get around it? So.

[00:17:13:17]Any more questions?

[00:17:17:02]I think I'm gonna stop here.

[00:17:22:07]We have any more questions, I think.

[00:17:28:04]Yeah, I think we covered we covered a lot. I had this whole.

[00:17:32:24]I had this whole notebook paper that like we've covered most of it.

[00:17:43:04]Yeah.

[00:17:44:15]It's it's one of those things that you kind of just you have to do it. And as

[00:17:49:14]you're doing it, you'll be like, oh my God, I can do this thing with it.

[00:17:52:21]And oh my God, I can do this thing with it. And you'll be having these

[00:17:56:22]conversations. And the agent will be like, oh, did you

[00:18:00:10]know that you can do this, this and this really quickly?

[00:18:02:22]And you'll be like, ah. So it's all.

[00:18:08:04]It's all an exploration. I like to say that this is

[00:18:12:07]the best strategy game you will ever, ever, ever play.

[00:18:15:10]So and you can go on any of our sub-site posts and

[00:18:22:08]ask questions, basically, like, well, we'll get it.

[00:18:27:09]Cassie or me. So what if you want to learn more about this and if you want to do

[00:18:36:03]more with like this was all the foundational stuff.

[00:18:40:19]Cassie, Alexander and I do run the marketing for storytellers sub-stack and we have

[00:18:46:08]weekly calls where we are going into depth on all of this stuff like this was an

[00:18:52:06]overview, but like going into depth and like digging in with all of our

[00:18:56:13]experiments, because we're always trying to find out

[00:18:58:13]how do we do things better, faster, smarter, all of that.

[00:19:02:16]So check that out. We do weekly calls on the paid sub-stack and we have at this

[00:19:09:04]point, we have a fairly enormous archive of calls and posts and all of these things

[00:19:15:13]and skills that we have made that we've shared that you can

[00:19:19:05]use with your agents as well and systems that you can build.

[00:19:22:15]If you like that command center that I was showing that I organize all my stuff in,

[00:19:27:16]I actually have a build out prompt on there that you can give to an agent and it'll

[00:19:33:03]start building that for you within your own systems.

[00:19:36:12]I know quite a few people have built their own with that prompt at this point.

[00:19:43:13]So if you want to go deeper, that's where you can go deeper.

[00:19:48:18]But as far as an overview, like just play with it. Just play with it. Have fun.

[00:19:57:07]It's really fun actually to learn how to have superpowers.

[00:20:01:18]But yeah, thank you for coming. I hope this was helpful. I really hope this was

[00:20:06:02]helpful because, and I know it's a lot of

[00:20:07:23]information so probably some of you are trying to absorb it.

[00:20:11:08]But yeah, this is kind of one of those things where go forth and have fun and be

[00:20:16:17]more awesome, right? Because that's always the plan.

[00:20:20:18]So thank you for coming and I will see you all around the internet.

[00:20:29:19]Yep. All right. Goodbye.

Notes

Topic times use MM:SS within this episode. Adaptable prompts are templates based on the lesson, not verbatim quotations. App names, options, and capabilities reflect the recording and may change.

Start with a usable context folder — 01:34

You do not need a complete studio system before starting. Begin with one consolidated place where the agent can find the materials for your work.

  1. Make copies of your working files.
  2. Gather the final versions of your books into clearly organized folders.
  3. Add images, character references, blurbs, product descriptions, and useful marketing material.
  4. Collect relevant reviews and existing social posts so they can be reused without gathering them again for every task.
  5. Add research and more detailed instructions as useful workflows emerge.

A database, Notion, or Obsidian can also hold context, but a folder is a valid starting point. Use the version of a book or description you actually want the agent to treat as current.

Choose a first task from your own workload — 06:48

Make a list of ten things that are tedious, repetitive, time-consuming, or that you dislike doing. Choose a task from that list instead of copying someone else's most elaborate system.

Possible starting points include:

  • Summarizing selected emails or newsletters.
  • Organizing an inbox using rules you have reviewed.
  • Drafting a newsletter from existing product information.
  • Preparing a batch of quote graphics or marketing material.
  • Checking a store for new products that need an established maintenance step.

Do the task with supervision first. An inbox-cleaning process, for example, should match your idea of what to keep before it becomes a recurring automation.

Build from one task to a system — 00:00 and 03:32

Do the task → improve the result → save the process as a skill → repeat it successfully → schedule it if useful.

A store-maintenance pattern is to perform an initial audit, agree on the desired approach, make the changes, and then check for new work on a schedule. The recurring job should apply the established rules rather than redo the entire project every day.

A marketing pattern is to combine book information, approved quotes, reviews, and images into a batch of assets. Once the process works, the same skill can generate another batch on a regular schedule.

A newsletter workflow — 07:57

Prepare these inputs:

  • A brief explaining the purpose of the newsletter.
  • The relevant products and links.
  • Previous newsletters that establish your voice.
  • Images, stories, or other material to include.
  • A repeatable skill describing your preferred structure and style.

The draft can then bring together copy, product information, links, buttons, images, and accessibility text in the newsletter system. Your remaining step is to proof the assembled newsletter and decide when it is ready to send.

Test repetition before a large batch — 09:43

For repetitive account work, such as the book-claiming task raised in the class:

  1. Check whether the agent can complete one item.
  2. Explain or demonstrate the steps if necessary.
  3. Watch several repetitions and correct errors.
  4. Confirm that it repeats the process correctly.
  5. Increase the batch size and continue checking progress and results.

One successful attempt is not the same as a proven repeatable process. Website behavior and available agent capabilities can change; treat a proposed workflow as something to test rather than assume it will work everywhere.

Give the agent a source of truth — 13:17

Before asking it to fix outdated public information, supply the correct information or identify where it should obtain that information. For example, provide the current edition details or point to your maintained book database.

Be explicit about the intended correction. Without that direction, the agent may choose a different online source and reproduce information you were trying to replace.

Adaptable prompt:

Use [current source] as the authority for this book's information. Compare it with [destination], identify the outdated fields, and help me update them to match. I will review the public-facing result.

Check the resulting page or record, especially when the change is visible to customers or readers. If the agent cannot complete a task, work through the specific obstacle rather than assuming an earlier workflow will always remain available.

Grow the system through use — 17:44

Keep the useful results, research, and instructions from each project. When a prompt or procedure repeats, consider turning it into a skill. The aim is to reduce repeated administration and make room for more creative work—not to build a complex system for its own sake.

For further study, the class points to AI Marketing for Storytellers on Substack, with deeper calls, posts, shared skills, and a publishing-command-center build prompt.