Claude AI Full Course (2026): Build Your AI Marketing Team with Claude Skills & Agents | One Shot
3801 segments
you've been using Claude like a smarter
search engine. You know, you probably
type a question, read an answer, copy
some of it. You're only getting about
10% of what it can actually do. So,
here's the thing, right? That's not a
prompting problem. It's not even right
about writing better prompts. It's about
having the wrong picture what the tool
can actually do for you. Now, most
marketing teams that I talk to probably
use plot the same way. They open a tab.
They type something like, "Write me a
campaign brief for our summer sale."
They read what comes back to them and
then spend the next hour turning that
text into probably a usable document.
Maybe a deck, maybe a tracker, a report
that manager can send to a client. Now,
you know, there's something about the
entire workflow that gets me going,
right? Because I know that the output
isn't bad. Claude's actually pretty good
at writing, but the process itself is
completely wrong because you've added a
step rather than removing one. So, what
I want to show you today is a different
way of thinking about this. So, not a
different prompting technique, but
rather a different entire mental model.
So, Claude isn't a writing assistant.
It's actually a workflow engine. And
once that clicks, every task currently
that eats up your time starts to look
like something you could hand off
completely. Not just partially but
completely. And this is the video and
this is the first video in a series that
I am calling flawed for marketing. Each
episode here covers a specific role. for
content writers, social media managers,
marketing managers, performance
marketers, brand managers and shows
exactly what changes when you stop using
Claude like a chatbot and start using it
like the workflow tool it actually is
meant to be. Today's video is the
foundation. If the reframe lands right,
everything else in the series will make
a lot more sense to you. So, let's get
into it. Okay, now let me show you
something concrete first because I think
abstract expla explanations of this
problem miss the point entirely. So
here's a task that most marketing
managers at least do once a week. You've
got a campaign brief, right? Let's say
it's for a brand called Orenda and this
is a midsize direct to consumer a D2C
skincare brand and they've got a Diwali
campaign coming up. Now, the brief that
they've given you is about two pages.
It's got the campaign objective, the
target audience, the key messages, the
budget, the timeline, the channels that
they're trying to activate, you know,
the standard stuff in the industry. Now,
your job is to turn that into a
stakeholder deck, something that your
brand team and the agency can actually
work from in the kickoff meeting. Now,
the chatbot version of this task would
probably go something like this.
Okay, so the chatbot version of this
task would probably go something like
this. So you would open claude, right?
You type something like here's my
campaign brief. So you have your entire
brief here, right? And you would say
turn this into a presentation outline.
So then you would just simply say that
hey you know uh this is my brief and I
want to create a presentation out of it
and then you'd probably mention for what
so let's say for a stakeholder kickoff
deck. Now you would want to also attach
the brief so I'm going to quickly do
that. So, I have the brief in the format
of
uh let's say a PDF. Right now, I'm going
to quickly attach this back into this.
So, let's say I'm going to do that and
then hit send. Now, I'm just going to
quickly wait for this to run.
So, I have the entire outline right
here. Now, what you get back is a text
outline. This is pretty good, right?
Actually, it's got a great logical
structure. So, it tells you, you know,
this is this is going to be on the
cover. That's the opportunity. This is
who we're talking to, you know, what
we're trying to say and the strategy at
glance, right? And it covers everything.
So, the campaign overview, audience,
messaging hierarchy, channel plan,
timeline, the thinking is absolutely
sound, right? But then what what do you
do with this? You probably just copy
this entire thing into PowerPoint or
Google Slides and you start building
slides manually. So you're formatting
basically text boxes. You're picking
fonts, adjusting spacing, adding a logo,
making sure that the title on each slide
are consistent. And probably let's say
an hour later, maybe 90 minutes later,
you've got even a longer brief that you
just saw right here is not something
that's broken, right? It works. And if
someone has been doing it this way for
years before actually Claude existed,
they'd say that Claude saved them time
because it actually did. It gave them
the structure. But here's the honest
question that I want to ask you. Did
Claw do the job for you or did it just
do one step of the job while you did the
rest? Well, the job was to not produce
an outline like this, but rather the
entire kickoff deck. What Claude gave
you was just an outline in the chat
window and you yourself had to sit
manually and turn it into a deck. So,
you're the person who did most of the
actual work and that's the chatbot
model, right? So claw drafts, you
produce, and it's genuinely useful, but
it's only 20% of what's available.
Here's the shift that I want to make,
right? Think about the difference
between two types of people you might
bring into this project. First person,
you hand them a brief, they write you
something, you take it from there,
right? And the second person, you just
hand them a brief. They come back with
the finished deliverable completely
formatted, structured, and ready to
send. Both are useful, but they're
completely different to work with. The
first one saves you the thinking time
while the second one saves you the
production time which is usually where
most of the hours actually go. Now cloud
for you can be either of those people.
Most marketing teams are just focusing
on using it as the first one. The teams
getting real leverage from it are using
it like the second. So what makes the
difference right? It's something called
a skill. So let's understand what
exactly is a skill. So a clot skill is a
preconfigured workflow. So it's not a
prompt, right? Uh you must understand
that a prompt is an instruction that you
type every time, right? A skill is
something that you only set once.
So once you have set it, you basically
end up using the skill multiple times.
So it's a trained workflow. So it goes
from let's say point A up till point B
then it goes to point C right and then
maybe it uses C to update B and so on
and so forth. So it's a it's a complete
workflow right for a specific job. And
here you usually have a trigger. So
something that invokes the skill. So
let's say an input right. So the trigger
determines when you need it, why you
need it, etc. So claude here uses the
skill and then knows the entire job. It
knows the output format. So it knows
what kind of a format
uh it needs to generate based on the
input that it receives. So the trigger
acts
as the point where you provide the input
and then it skips has its own number of
steps. So it goes from A to B to C to
produce an output that is helpful for
you. Right? So you give it the input and
then you directly get the output or the
finished thing. Now think about what
that means for this Diwali deck example.
Right? So instead of writing a prompt
something like this and asking claw to
give you the presentation outline, what
it does is that you upload the brief,
you trigger the PowerPoint X skill. So
this skill could be something that's
already inbuilt and it could also be
something that you create. And
throughout this series, we're going to
do both of them. So in today's video we
will actually invoke a skill and in the
future ones we'll see how to write them
on our own as well right. So here we're
going to invoke the PBTX skill and plot
doesn't give you the text right. So for
at the end it'll actually give you a
PowerPoint file you know structured
slides headings content blocks and
probably a format that can directly open
in PowerPoint or Google Slides and
that's the difference right so you have
the same input but instead of actually
using a prompt
what you end up doing is that you use a
skill and then an output is generated
from that skill. And now imagine this,
okay, this is not just limited to
presentations.
There are skills for content writers,
for social media managers, for
performance marketers working with PDF
reports, for brand managers who are
building documentation. Each one is
trained for a very specific role and a
specific kind of deliverable.
And we will cover each of these in
dedicated videos in the series. But
today I actually wanted to show you the
comparison right with the same brief. So
the brief that we're going to be using
is the same brief that is here and we'll
see how it differs from the chatbot
version and the skill version. So the
different isn't abstract cuz you can see
it right here.
All right. So this is the Orenda
skincare brief that let's say you
received right. So this is a Diwali
campaign brief. uh it's about three
pages long and this is the kind of
document that actually lands in a
marketing manager's inbox right and it
needs to be created let's say into a
stakeholder deck by the end of the day
now before we had seen how to do this
using the chatbot version which is
something that you'd probably do every
day and how most marketers also function
right so you go here you attach the file
maybe you know you give it the brief a
little prompt and then what comes back
is an outline, right? And this is a
structured outline for the kickoff
there.
Now, um so you know, I'll probably be
honest, this is what most marketing
managers are doing. The the output is
pretty good and claw going to also give
you back something that's logical,
right? So, this is solid. you've gotten
an executive summary and each and every
outline, a complete outline and each and
every aspect that you actually require,
right? And uh but
this is the whole point, right? You're
right now just looking at
text in a chat window. To get a deck,
you have to open PowerPoint, start a new
presentation, build this slide by side,
you know, copying one by one and making
it look like something that the client
would actually like. And you know, I've
timed myself actually doing something
like this last week, and it almost gets
me 2 and 1/2 hours to go from this to a
usable 12 slide to, you know, 15 slide
deck with just working, you know,
quickly and having no design tweets. So,
let's do the same skill with the
PowerPoint presentation skill, right? Uh
let's figure this out how to write here.
But instead of actually writing a prompt
directly, I am going to invoke a skill.
So how am I going to do that? I'm going
to trigger the PPTX skill and get it
done. So here's how I'm going to do it.
I'm just going to quickly type out and
uh say use the PPTX skill create a
stakeholder kickoff deck from this
campaign brief.
So now I have attached it right let's
say uh instead of highq maybe let's use
sonnet the model for this particular
uh task and now let it run and you know
let's just wait for it to run
okay so what's happening right now is
very different from what had happened
before right so claude isn't generating
an actual outline for me rather it is
creating a complete PPTX so a complete
PowerPoint presentation for me. So it's
making the actual file. So you can see
here it says creating the kickoff tag
and if you go down you see that it's
writing some kind of code and you know
it's installing stuff and it's creating
the outline for uh not just the outline
but it's also creating the actual actual
campaign deck that I require. So you can
see things like font size, you can see
things like shapes etc. And now it just
says generate the PBTX. So it looks
solid now. visual things. It's adding
some images into it. It's finding some
other scripts and then it's building it.
Okay. So, this is what came out, right?
Um, my tool had actually hit the limit.
So, I had to press continue so that you
know it can continue the uh entire
process. But now I have this right here
in front of me. This is my complete deck
that I need to send. So, instead of just
having, you know, uh the document
outline and me having to put all of it
into this, now I've got my title slide,
my campaign name, my campaign dates, a
summary of what's going to be on today's
agenda, I've got the objectives, I've
got the entire overview, what's the
budget, what's the campaign window, you
know, when does it go live, each and
everything. And all of this probably I
got it under 3 and 1/2 or maybe 4
minutes. You see the difference? I am
saving hours here. This is a 12.12
slide deck and it can be opened in
PowerPoint. It can also be added to your
Google Drive. The formatting is
absolutely clean. Each and everything is
from my PDF that I had uploaded.
Everything is readable. There are some
touch points that they have mentioned.
There is allocation with intent that it
also shows. It tells me what we're going
to produce the kind of critical
milestones you know for the campaign to
be live for 21 days to own Diwali.
Now from just uploading the brief right
to having this file I probably just took
5 minutes. Five. That's all I needed.
Five. Now, think about this this way,
right? The only difference is whether
Claude
gave me text to work with versus it
actually gave me the finished thing. And
that's the skill model. And that's what
I mean when I say Claude is a workflow
engine and not a chatbot. It's not doing
one step of the job and handing it back.
It's actually completing the job for
you.
So the question now at hand is where
does this apply in your week? Here's the
way that I like to think about it,
right? Take any task where you currently
use CLA to generate the text and then
you spend time turning that text into a
deliverable. That gap between Claude's
output and what you actually need to
send, that's where a skill lives. Okay.
So for most marketing teams, these tasks
actually fall into let's say three
buckets. The first bucket that we have
at hand is called content creation. So
this is the bucket where you're writing
copy script, social posts, emails. So
you generate something in claw, then you
spend time cleaning it up, removing the
air tone, formatting it for the
platform. The humanizer skill actually
handles that cleanup for you. the
caption generator, the real script
generator, those handle the output
format directly for you. The second
bucket that is at hand is the bucket
that is related to data and reporting.
So let's say you're getting PDFs from
agencies, right? You're pulling numbers
uh out of reports, rebuilding
spreadsheets manually. The PDF reading
skill extracts and structures that data.
The Excel skill will basically build a
spreadsheet for you. And then there
exists the third bucket which is
presentations and documents. Now this is
something that I just showed you here.
So presentations and documents creates
your decks, your brand guides, your
client reports, everything. So a PBTX
skill, a DOCX skill, a PDF output skill,
all of it can get covered within this
little framework that I just gave you.
So if your team does any of those
things, right, and most marketing teams
probably do all three of them, there's a
skill that can change that workflow. So
here's a practical thing that you can
actually do um you know this week. So
let's say think about any one task
right. So I'm not asking you to think
about multiple tasks. All I'm asking you
to think about is that one task that
takes most of the time to go from let's
say the claude's output to something you
can actually use. And that's going to be
a starting point and that's also going
to be the first skill that you can start
learning properly. And it's also going
to be oh it's also going to be worth the
first skill that you learn here. Now
I've linked the relevant episodes in the
description below. Each one covers a
specific role and goes deep on that
exact workflow. So find your role and
probably you can start right there.
Here's the oneline version I keep coming
back to. Right? A skill is basically a
trained junior in your team for one
specific job. So you don't manage them.
You just give them the brief and they do
it. So that framing matters because it
changes how you use the tool. You
wouldn't hire a junior team member and
then stand over their shoulder the
entire time rewriting everything that
they do. You would give them the brief,
let it run, and then review the output.
And that's exactly how you use a skill.
I've put down all the differences for
you. So it's easy for you to understand
in a tabulated format. And let's quickly
run through it. So here's the difference
between a prompt and a skill. So what
does a prompt do? In a prompt, you have
to write the instructions every single
time. So every time you want to do
something, you have to write it again.
You describe what good looks like. So it
varies. In a skill, on the other hand,
the instructions are preset. usually or
when you develop your own skills, you
set the instructions yourself so you
know what you want every single time.
Then in a skill, the quality standard is
usually built in. So this is especially
in relation to let's say skills that
make PowerPoint presentations or
documents or Excel files. These are
inbuilt skills. So in inbuilt skills,
the quality standard already exists. And
when you're creating a skill, you can
also set the quality standard. Now, a
prompt, the output is usually text in a
chat window. And um in a skill, the
output is usually a fixed
a finished file or a deliverable. These
deliverables are known as artifacts in
the clawed ecosystem. So these artifacts
can be HTML pages
right or SVG files depending on what
your output is and how you describe it.
Um sometimes prompts can also lead to
the creation of a deliverable but it
depends on what and how you exactly
define the output. In a prompt, you have
to do the production work yourself
because there is no end output. It's
just text that is given to you. So
translating it into a PowerPoint,
um you have to do all of that production
work yourself. In a skill, the
production is part of the skill, right?
Prompts are inconsistent. Basically, it
depends on how well you prompt it. A
skill is consistent because once you've
set the standard, it runs the same way
every single time.
So the last row is actually the one that
matters the most at a team level and
we'll come back to it later in this
video.
All right, let's get on to the practical
aspect. So we understood the
differences. Now let's get to the
practical part. So I've opened Claude,
right? This is what the interface looks
like. You might be familiar with the
chat interface. So it says good
afternoon chhatanya that's my name and
here you have different options to
either write something to create
something to create code bring something
from my calendar and bring something for
Gmail. Now these two are connectors.
Basically what a connector is is it
connects with your claw system
uh to give access to any kind of
external apps. In the claw interface you
also have different models that are
accessible for you. So hiq gives you the
fastest answers. Sonnet gives you the
most efficient ones for everyday tasks
and opus 4.5 is the one which is the
heaviest model. So this is something for
let's say developing a website. I would
probably use the opus 4.7. Now um I'm
going to quickly switch to highQ 4.5 cuz
most of the formatting stuff uh can be
done by IQ 4.5 and it's very low when it
comes to token usage. Now what is a
token? A token is essentially the
currency that Claude uses to generate
answers for you. So you would want to
limit your tokens because you don't want
to exhaust through them quickly. Now up
here in this interface when you go here
which says add files, connectors and
more here you can see something called
skills. Now I have created a lot of
skills. I have imported a lot of skills
that I have found on the internet. These
are skills that I use on a daily basis.
Whether it's an AI weekly digest or it's
a canvas which creates designs and PDFs,
whether it's content writing, content
research, whether it's creating a new
skill in itself. Now, the skill creator
is an inbuilt skill that comes include
which lets you create skills for
specific use cases and that's what we're
going to also use in this video to
create a new skill.
Now u there are certain skills like
PPTX,
XLX
and PDF creator that are inbuilt skills
which you cannot access from your
dashboard here. So if you go to manage
skills, this is what opens. So these are
all my personal skills if you see right
and all of them uh humanizer is a very
famous one that's available on the
internet. I recommend you use this uh so
that you can remove signs of AI
generated writing from your text. Now um
you can also see skill creator is a
skill where you can see the method that
Claude uses to create a skill for you.
Similarly here you can see the actual
markdown file or the code. So you can
toggle between the code and the preview.
This is the preview mode where you can
see what exactly it does. So it gives
you context on how to create a skill,
how to write the markdown file, you
know, what are the writing patterns to
follow, what is the output format, what
are the examples for the output format,
the kind of writing styles that are
required in creating a skill, everything
comes from here. So what are the use
cases to create new skills? Modify and
improve the existing ones. Measure skill
performance. Use it when users actually
want to create a skill. Edit and
optimize an existing skill. Run
evaluations to test a skill and
benchmark skill performance with
variance analysis or optimize a skills
description for better triggering
accuracy. This description is the part
which actually helps you to decide or
even cla to decide when to use a skill
in chat. And this is probably one of the
most important ones, right? And if you
see the skill creator, it says added by
who? Added by anthropic. And the trigger
is usually the slash command. That is
you do slash skill creator or
automatically whenever it's relevant.
Now if I go to some other one like
caption generator is something that I
did. So it says added by you when it was
last updated and how it's triggered. Now
let's get on to creating and utilizing a
skill to make a PowerPoint. Okay folks.
So what we're going to do right now is
utilize the inbuilt skill to make the
PowerPoint presentation using the orinda
brief that I covered in the last
episode. Right? So I'm going to quickly
write help me make a stakeholder
presentation
for
proposing or for this brief.
Very simple. And uh let's see if H HighQ
does it or I'll have to maybe change the
mode to set. Now it's going to read the
campaign brief. It's already done that.
The next step what it's going to do is
read the presentation. So it's reading
the PowerPoint XKL. So I want you to
understand this. See, it's reading the
PowerPoint XKL. Now it's installing the
required scripts for the presentation
right and then it's running the command.
The running the command is the command
that it's running to make the
presentation itself.
Now while that's running this is the
moment where most people's instinct is
probably adding more instructions. Um
you have to be absolutely patient with
this. It takes time to run but it
delivers one of the best results. So if
you have the urgency to write something
like make sure the slides are clean,
keep it under 12 slides, use a
professional tone, you actually don't
need to. The skill already has those
standards built in. And that's usually a
sign you're trying to override. If
uh you tell it not to do it, it'll
basically override the skill if you
prompt heavily inside the skill. And
that's a sign usually where you're
trying to override it rather than using
the skill. So take a seat back and just
relax. And if you see it's running
commands, it's editing the slide. It's
verifying the presentation. It's doing
everything. Now it's giving me the
entire presenter guide along with the
deck.
Now if you see it's presentation
generation, converting presentation to
PDF for image interaction, converting
PDF to individual slides. So it's doing
each and everything. See, it's doing
visual inspection.
It's verifying the final presentation
slide. And now it's creating the
presentation. This is a markdown file
using which I think it'll create a
presentation uh for this. And now the
PowerPoint is here for you. So if you
see here the entire PowerPoint is
available along with a presenter's
guide. So a presenter's guide is
basically telling us about the talking
points of what decks this deck is not
and how to present it. And this is the
stakeholder Diwali deck that will be
utilized
for
yeah that will be utilized for this. Now
I used highq so the kind of colors that
they use and everything is a bit um you
know basic but let's see what happens
when you upgrade the model with the
skill. All right folks so now I'm going
to try and upgrade the model to sonnet
4.6 six and I'm going to quickly add the
file the orenda brief right here and I'm
going to simply write create a
presentation
for
take
holders
based on this brief. Let's see how this
model performs. Um
see again it says it's reading the PPTX
skill. So this is the PPTX skill
actually. Let me quickly show you this.
Uh by the time it is going through
everything and creating the presentation
so it's setting up dependencies and
doing all of that. So in here you see
the description. It says you can use
this skill anytime a pro point X file is
involved in any way. Input and output
both. Uh then it says what to make,
whether to make, how to read and analyze
the content, how to edit and create from
a template or create from scratch. And
then it gives extensive guidelines of
what it should do. It also says what not
to do. So what kind of mistakes need to
be avoid uh avoided in creating the
actual PowerPoint. It talks about
textbox padding. It talks about do not
default to cream or beige backgrounds.
think about these different colors or
users brand palette. Now, this is also
something that you can set inside um
your claude which I have not yet but I
will show you how to set that up in
upcoming videos as well. So, how to set
your own brand palette
and then there is other things about
visual the different sub aents even for
two three slides. Then it gives code and
it even talks about converting to
images. Now you cannot save this skill.
If you see I will tell you why. Because
the skill name is already in use. So you
already need to have another thing. Why?
Because this is an inbuilt skill
include. So there's no way that you can
redo it. Let's check where our
presentation is. Okay. So it is at the
part where the Java script so JS uh
script is being written. So you can't
see the file yet for very obvious
reasons. But let's also understand what
language is the skill written in. The
skill is written in
a markdown language. So this is
markdown. This is another pro uh I
wouldn't say programming language but
another document writing language. All
right. So this is completely created and
it also hit its limit. Uh this happens
more often than not because if there are
any issues it also tries to resolve the
issues. So there were two issues and
then the tool usage limit hits. So you
have to basically click continue so that
everything gets
delivered to you. So it's doing
everything. So it's it's it's literally
trying
to create
a well rendered and complete file that
is with the correct output. So now it's
rechecking everything ensuring that
everything that's given to me it's
perfect. And now it says that all the
slides are clean. So it's copying to
output and there you have it the entire
Orenda Diwali campaign created for us.
Brilliant. And you see the moment that
we actually upgrade
our
uh model, the kind of campaign, the kind
of designing also goes better. So
designing can even go better if I change
to opus 3 or even opus 4.6 and uh yeah
it'll basically create a brilliant
PowerPoint like this. Now we saw this in
the last video right? So this is very
similar to that. But now let me take you
through the next section which is
actually understanding what kind of
skills to invoke when you are working in
a different role inside your team. So
the question I get most often at this
point is how do I know which skill to
use for a given task? And there are two
ways to come at it, right? And both of
them are valid. So you can either choose
by your role. So say I'm a content
writer and what applies to me or you can
choose by output you need. So I need a
spreadsheet. So what skill produces
that? So let me walk you through uh
through that both of them in the next
segment.
Now let's first understand what skill to
use when according to the different
roles. Right. So, a content writers,
your main skills are going to be
something like a humanizer and the dog
X. So, humanizer strips the AI tone out
of copy and makes it sound like an
actual person wrote it. On the other
hand, the dox skill takes your content
and makes it into a clearly formatted
document that can be sent ahead to
either clients or to the various
stakeholders that require this copy,
right? Whether it's for a website or if
it's for a real or if it's a blog,
whatever it is, you can send it ahead in
a properly formatted document.
For social media managers, caption
generator is a very simple skill that
you can create using the claude skills
create especially the skill uh creators
the skill creator skill. Wow, that's a
mouthful. But the skill creator skill
and it takes a script or a topic as an
input and it makes ready to post
captions according to the different
platforms that you need to post it
across. A real script generator is
something that I made to help me write
and produce short form video scripts in
a specific length and format. So
whatever suits me, that's the way I
write it. Both of these get full
episodes in this series. So don't worry
about it. We'll cover both of them
extensively.
Now for marketing managers, right, what
is it that is your biggest headache?
That's creating decks and making
trackers and budget sheets. So there is
a PowerPoint X deck uh creator that we
just saw in action. Similarly, there is
an Excel file creator and reader. So it
handles your data. It creates structure
for you and gets back a working
spreadsheet in the right format and this
is important if you're somebody who
handles data analysis for your marketing
team on a daily basis. Now a performance
marketer again XLS the Excel file is
something useful because a lot of your
workflow also involves PDFs. So reading
PDF reports from different agencies or
platforms, right? Whether it's invoices
and the PDF reading skill is
specifically designed to extract and
restructure this kind of data. And once
you pair it with the Excel skill, you
have replaced the copy paste grind from
all the agency reports. Now
the part that is important is the brand
manager skill. So you can have docx for
guidelines that you would want to create
to tone documents to create briefs. You
have PDFs to make client ready reports
and deliverables and you can create your
own brand voice skill that can help
decide the formatting of all the
documents that you have to send ahead
and all the creatives that need to be
put out. So according to role these are
the skills that we think are going to be
useful for you. Now let's go to output
type.
All right by output type there are a lot
of things that skills can do for you. So
if you know what you need to produce but
you're less sure which skill hand
handles it, you can work backwards. So
for example, if you want to create a
PowerPoint,
you can use the PowerPoint skill. To
create a document, you can create a DOCX
skill, right? And if you want to read
Excel files, you have this, right? If
you want to understand a PDF that is
read it or write a PDF you have the PDF
reading and also the PDF writing skill.
Now all of this is there if let's say
you want to humanize data right that is
humanize your writing you have the
humanizer skill. If you need a script
then you have the real script generator.
If you need a platform
uh specific caption generator, you have
that. Now, all of this is available, but
there's one honest caveat here. That is
some tasks may span across skills. So,
what does that mean? It basically means
that instead of one job being fit into a
particular box of a skill and pretending
that every job fits neatly in one box,
we are trying to understand that all of
these can work across each other. So let
me give you a quick example. Let's say a
campaign report comes in and it's in a
PDF format, right? And then it needs to
be written in the form of a docx, right?
That your team can edit. So it is going
to touch two skills. One PDF reading and
two dockets to create the actual dock.
We cover those multi-kill chains in the
later episodes. But for now, if you're
just getting started, pick one skill,
one output type, and get the pattern
down. Then you can start chaining
better. All right. Before I get into the
demo and the practical side of things, I
want to actually spend 2 minutes naming
the problem specifically because most
people know AI copy sounds off, right?
But they can't always tell why. And
knowing why makes everything that
follows make so much more sense. So
there are four patterns that show up in
almost every piece of AI generated copy.
Once you can name them, you'll probably
spot them in 3 seconds. So if you look
towards here the first is vocabulary. So
there are set of words and phrases that
AI uses constantly and almost no human
writer does whether it's writing
something like delve into it is worth
noting X underscores Y in today's
rapidly evolving landscape right instead
of writing in today's landscape or
something like pivotal moment. If you
see any of these, you're probably
looking at an AI copy. So, beware of
that. Next is sentence rhythm. So, what
exactly is sentence rhythm? Well, the
easiest way to understand this is
actually this graphic that was given by
Gary Pvost in his 100 ways to improve
your writing. This is an excerpt for
that from that. And if you see here, if
you read a paragraph of AI writing and
count the sentence lengths, right?
They're almost identical, whether it's 8
words, nine words, 10 words, 8 words.
Well, you can see it something like
this. So, this is how the excerpt goes,
right? This sentence has five words.
Here are five more words. Five word
sentences are fine, but several together
become monotonous. Listening to what is
happening. The writing is getting
boring. The sound of it drones. It's
like a stuck record. The ear demands
some variety. You see how monotonous
that can get, right? But you know how
actually humans write? Human writing
varies. A short sharp sentence maybe
followed by a longer one that takes its
time getting somewhere. Maybe changes in
directions partway through lands
differently than you expected. Right?
And AI writing is metronomic. So that's
what makes it feel very flat even when
the information is right. If you see
here uh what Gary Pos says is now listen
I vary the sentence length and I create
music
the writing sings it has a pleasant
rhythm a little a harmony I use short
sentences and I use sentences of medium
length and sometimes when I am certain
the reader is rested I will engage him
with a sentence of considerable length a
sentence that burns with energy and
builds with all the imp impetus of a
crescendo, the role of the drums, the
crash of the symbols, sounds that say,
"Listen to this. It is important." So,
write with a combination of short,
medium, and long sentences. Create a
sound that pleases the reader's ears.
Don't just write words, write music.
Well, it's a brilliant piece of writing.
Um, but I think you've understood what
sentence rhythm is from this. Now, the
third aspect that AI covers is actually
inflation, right? So what does that
mean? AI has a specific habit of ending
paragraphs by stating that whatever it
just said is important. So this
underscores the significance of X for
brands seeking to Y. The sentence adds
no new information. It just points all
the previous sent it just points at the
previous sentence and says that that
sentence mattered. See humans don't need
to write like that, right? We trust the
reader to decide what mattered and what
did not matter. And the fourth rule in
this entire fiasco is actually the rule
of three. Now AI defaults to distinct
things in groups of three, right? So
whether it's apple, bananas, oranges,
cars, bikes and trucks, eating,
drinking, dancing, everything, right? So
without variation, it just goes always
in a rule of three. You'll notice it
when you're reading a piece and every
single point comes across in a trio. So
that's usually actually a finding its
comfortable pattern rather than a writer
making a structural choice. Um now that
we understand the background right the
humanizer skill by a blader is actually
trained specifically to catch all four
of these. It's not a style preference
tweak rather it's a systematic pass on a
documented list of AI writing patterns.
It's a moment I'll show you uh where and
how to do it. So in a moment I'll show
you exactly what it does to a real piece
of copy. So let's get into it now.
All right. So you can see here I'm going
to use the same brand as Orenda last
time. Uh the one that we've used across
our series and I'm going to ask clots.
This is standard chat interface to write
a blog post. So same brand and I'm going
to ask to write an article that's aimed
at marketing managers at D2C. So if you
see that the audience that's marked here
is marketing managers at D2C brands. The
tone that I want to pick is
authoritative and it should be titled
five reasons why D2C skincare brands are
losing customers in their 30s and what
to fix. Let it run. This will uh rightly
do it. So right now it's picking up my
content writer skill. So this is a skill
that I had developed to write content
right uh to write exactly you know what
I want and how I want it. So it's
writing it from that perspective. Now
uh let it's quickly generated the post
and you can see it here. So it's given
five exact points to fix. So your
product scale claims don't scale with
maturity. Your pricing strategy assumes
loyalty that doesn't exist. You're not
building this. Your content strategy
treats aging like a taboo. You're
chasing Tik Tok trends instead of
building a habit. And it quotes a real
problem. Right? All of this again is
written using the content writer skill
that I have. But this is still not
humanized. So how will I use the
humanizer? Right? So the humanizer skill
is actually not going to change the
structure of the entire blog that's
given to me. It's going to only change
the language. So let me do one thing
quickly. Invoke the humanizer skill and
say humanize
this
blog for me.
So now it's going to humanize this blog
and it's going to understand what all is
required to be rewritten. And now let's
quickly see the difference between the
two of them and we'll understand this by
placing them on a word doc. So I'll
quickly do that and then I will explain
it to you. Right? So first it's taken
the entire part of the rewrite and then
it says what obviously makes the below
so obviously generated. So words like
this isn't random, it's systemic. The
artificial triplet rhythm random,
systemic, expensive that feels
constructed rather than discovered. It's
costing you serious money. It's a vague
threat to maintain engagement. Here's
where you're falling, right? a bulleted
list setup and fix it feels like formula
like
the significance aspect that we talked
about right the overestimation the
entire value proposition breaks down
overestimated significance even heavy
use of m dashes short punchy so
basically it tells you what all is the
problem and how to fix it and then it
fixes it for you so it starts off by
saying your acquisition engine works
You're pulling in 22 year olds with
Instagram ads about dewy skin and 0.5%
retinol, but somewhere between year 1
and year 3, they're gone. The customers
who found you at 25 don't come back at
30. And unlike losing them to a
competitor, this one's quiet. They just
stop showing up. It's predictable and
it's expensive. The D2C model was built
for fast product cycles. Trends come and
go. People swap formulas constantly.
novelty drives purchases. Then your
customer turns 30 and the whole setup
stops working. She's not chasing the
next viral serum. She wants something
that actually works and it won't deliver
it. She'll find someone who will. So
once you don't deliver it, she'll find
someone who will. And then after that,
it gives an entire thing that's already
humanized and it'll tell you what kind
of changes it made as well. So that's
how you use the humanizer skill. And
this was just a run through of how you
can use the humanizer skill. Now let's
turn this copy into something that
someone can actually open. So let's uh
get on with the next segment that is
using the docx skill.
Okay, back to this now. Now we're going
to make this into a dog. So I'm going to
quickly write create this into a
properly formatted doc for me and I'm
going to save this here. And now it's
going to basically take this entire
thing. It's going to read the docx skill
to understand word document formatting
best practices and create the word
document for me. Now uh there is you
know maybe some kind of formatting notes
that you would want to add. I have not
added those yet at this stage. Actually,
I should have already done that. So, how
about we go back here and write some
formatting notes. So, let's say blog
post format
H1 for
the article
title.
Article title. Let's say uh H2 for each
of the five
five
section headings,
body text
for everything
else
and full
article
content.
Okay. Uh there are some typos in this
but please bear with me. But now it's
going to create an entire thing that's
properly formatted with the blog post
using the docx library. It's going to
use each and everything that is required
and it's also going to take reference of
the little uh you know notes that you
have added. So I've added a formatting a
brief formatting note right and that's
all it needs. So let's wait till it's
generated and let's see how it turns
out.
All right folks, you see this? This is
what it is. Now I have the entire blog
in a docx file that's properly formatted
with each and everything. So this is in
H1, the five D2C brands. The five
sections as you can see is H2
heading. You can see the document
outline on the left if you're in Word.
And the body text is clean throughout.
There's consistent spacing. There's no
inherited, you know, formatting issues
from pasting into a blank document. Now,
if I send this to a client or a brand
team today, right, this is ready. This
is the final read through. Maybe one or
two tweaks and then directly goes out.
So from a brief that was here that is I
want to just write a blog post all the
way up till here what we've done is
create an output that is uh that is an a
dox file right so now how do I apply
this this to my ecosystem is I'm going
to help you create actually a content
writer skill. Now, if you see here, I
already have a skill that's a content
writer. So, this skill drafts fulllength
marketing and business content
specifically for posts, articles, etc.
Now, let's say I did not have that
skill, right? So, how can I acquire this
skill? Well, there are two ways to go
about it. One is to search for it online
and the second is to develop it
yourself. So let's say I use something
like a skill creator and I'm going to
ask
it to write content uh for let's say a
particular task. So I want to create
let's say
help
me create a skill
that
writes
blogs
on
let's say makeup products that's the
industries I work in right on makeup
products
as a D2C
C brand
okay so now it's going to involve poke
the skill creator skill and it's going
to help me make a skill for this
particular thing. So let's say writing
makeup product blogs let me first upload
and understand the best approach. Now
they already have this and now it's
going to ask me some clarifying
questions, right? So let's quickly
answer the clarifying questions. The
first question that it's going to ask me
is what is my primary use, right?
Whether it's to drive SEO traffic,
whether it's to build brand authority or
product education. Let's say I am going
to look at product education, right? Um
how much product research or context
should the skill assume Claude has?
Should it product from your actual
product data or work generally
generically for any model product you
name? Let's say I will go here work
generically and I want it to be medium
form. Right? So now I have given it all
the necessary inputs and the parameters
that it requires to build the skill and
now it's going to ask me some more
questions about my brand voice. So let's
say I want to make this inclusive and
empowering and I want to also make this
more stepby step right and uh the
product page copy should be more of an
educational blog of how to actually use
the product rather than make it
outrightly uh you know
outrightly in your face to buy the
product and very CLC. Now, what's my
preferred blog structure? Well, should
it be lean, direct or include contact
sections? Let's say I want a hook and
yeah, as simple as that. And yes, always
include SEO. And what should be the
assumption that is this?
Yeah. So, I've just like picked up some
things and now it's going to draft my B
D B D B D B D B D B D B D B D B D B D
B2C blog skills. So, product blog skill
like this. And you can see what it's
actually doing, right? If I open that up
and while this is going on here, uh let
me quickly
get back to this once it's done.
Okay. So if you see it's also doing some
test prompts on the skill. So it's taken
some test cases. It's verifying if the
skill is actually readable and it's
taken a test prompt uh as a foundation
and now it's actually writing it of how
to apply foundation on your face. So it
says skin blurring foundation a
step-by-step guide. This is the meta
description. This is the target keyword.
you know what it does actually etc. And
now it's also doing a skill summary and
quality assessment of whether the blog
that it wrote is actually good or not
and if it does something that's
important um in terms of SEO or not. And
once it's done it's actually basically
testing out all the different products
that you know uh can be used. So it's
also testing it out on a different
product type. But this is how you make a
skill that's particularly for let's say
your
use case. So here I have taken the use
case of makeup products. The one before
that I taken the use case of skin care.
You can use it for any segment that
you're in. And this is how to make this
more concrete rather than just something
you watch right. So I want you to find
uh uh find a use case in your particular
brand where you have to do a lot of
writing. So you find that one piece of
AI generated copy that you've already
edited this week. Something you ran
through claude and then manually cleaned
it up. You find the original draft and
the end draft and then you can run that
original through the humanizer and
compare the output of what you actually
understood uh what you edited yourself
and what you needed
the humanizer to edit for you and
compare both of them. So meanwhile I've
seen the makeup blog editor is ready and
it's given each and everything checklist
when to use it when not to use it right
the kind of iterations required why it
feels off you know the quality checklist
etc everything is ready now you can
simply save the skills I'm not going to
save the skill because obviously I am
not you know using it for any particular
um product cuz I don't have a makeup
company but if I were to I would
probably look at this practical skill.
Okay, before the demo, I just want to
spend about 2 minutes on why generating
captions in a plate cla chat isn't going
to solve this. So when you ask Claude,
right, with no skill, just in a standard
chat to write you an Instagram caption
and maybe also look at a LinkedIn
caption for the same topic, what you get
back actually is just the same copy but
applicable
in just two different rappers. So
instead of having entirely different
things that the same text can focus on,
um it just basically says
let me change the language and hope that
it works for both the platforms. Right?
So it uses the same sentence structure,
the same tone. It's roughly the same
length with a few words trimmed here and
there. and Claude doesn't know by
default that Instagram and LinkedIn are
actually read completely differently
because we all know that Instagram is a
platform where everything is a bit
casual but on LinkedIn you have to be
much more professional in nature right
so that's the difference so Instagram is
probably for somebody who's just
scrolling right the thumb is moving and
you have about half a seconds to stop it
the caption that works there would
probably be something that's thought,
right? And that's very specific, right?
A statement that creates maybe a jolt of
recognition or curiosity before they
even decide whether they want to keep
reading or not. And then it earns the
rest of the post. Now on LinkedIn is the
other way around. The person reading it
has paused because they actually want to
read the post. They're in a professional
mindset. they're reading it for the
relevance part and they'll give you two
or three sentences before they decide
whether they want to continue reading or
whether they want to discard it. Now the
pose that works here probably opens well
with an insight and not the hook. So it
assumes slightly more patience and
slightly more context. So it's the same
idea with completely different opening
moves and that's the difference. So
which platform content writers develop
an instinct for over the years is what
the caption generator skill is going to
be trained on. So it just doesn't
reformat the copy. It actually rewrites
the reading context of it. And we're
going to do that for Instagram LinkedIn
X because this has again a character
limit of about 270 characters, right? if
you don't have the premium and more uh
if you do have the premium version and
YouTube let's say you are posting one
video across Instagram, YouTube,
LinkedIn and X how is the caption going
to vary then that's also something that
we're going to look at. So let right so
let me show this to you and let's get
right into it.
All right, folks. So, let's begin with
actually creating a
skill
for a real script generator. Now, this
is going to be a real script generator
that can also work for YouTube shots.
So, basically I have two aims. One
taking an idea. So it takes an idea and
generates
two scripts.
One I'm going to say YouTube shorts
and one for let's say Instagram res.
Right?
The focus
should be on
generating
scripts
in a hook
content
CTA style. So I've given it a style as
well. I have a hook, then I have the
content, then I have the CTA style
for Orenda
Skinare. Okay, I am attaching
a brief
of a campaign
that we did some time ago.
Please take a look and then ask me
clarifying
questions
for the same. So now it's going to ask
me clarifying questions as well. I'm
going to quickly change to sonnet so
that it does better thinking and then
we're going to go here give it more
options and within 5 minutes
approximately
I am going to
oh s me I forgot to actually attach the
orenda brief so it's going to say seems
like you haven't uploaded the file yeah
that's all right so we have uploaded the
file Now I've added the Orenda skincare
brief as well. It's a solid Diwali
campaign. So now it's going to ask me
some questions. Uh so you see that the
questions it's asking me are very
specific. So instead of asking these
questions like this, I'll be like ask me
in MCQs
where I can select because I'm lazy. I
do not want to type anymore. I can even
ask it to ask me in an MCQ format where
I can just directly select. So now it's
going to ask me in my questions in a MCQ
format that's an interactive artifact
that comes out of it. So
what's my brand voice? So while this is
generating
uh let's just wait while it's going on.
Okay, so it's asking me questions now
about so it's generated all the
questions, right? And let's say let's
quickly pick these out. So what is
Renda's brand voice? Let's say it is
more
aspirational and premium kind of script
length I'm targeting. YouTube shots is
going to be 30 seconds. Instagram res
are going to be 15 seconds. What CTA uh
drives viewer? What should the CDA drive
to view? Um, I'm going to say shop for
links and maybe also visit our website,
right? And save the video because that
generates engagement. Let's say I am
going for CTA differ between. Yes, they
do differ per platform. Which product
should the script cover? Uh
let's say skill asks user each time name
products mentioned
in this script.
Uh maybe you know let's make that answer
hook style should the skill default to
question based bold statement generate
all three and maybe let me pick one
should YouTube shorts and Instagram res
feel totally different yes
that is true any claims that are off
limits no competitor mentions let's say
I want to say stuff like anti-aging cuz
that's too much in the market and also I
want to make no medical claims, right?
Um cuz legality problems. Now all of
that gets spilled into it and now the
skill creator skill is going to
understand how to build a proper skill
and it's going to build me a real script
skill. Now I did this specifically for
Orenda using a campaign brief that I had
used in my previous videos. But what you
can do is you can make one for your own
brand, the brand that you are looking at
or you could make it campaign specific
by saying that every time I add a
campaign brief or I attach a campaign
brief with the idea of I want to make a
real hey give me invoke this skill that
gives me an overall line and structure
of what can be done. So now it is
creating that real script generator. So
if you see it's writing the entire
markdown here
and I'm going to quickly wait for it to
be done so that I can show this to you
and then I'll save this as well and I
will invoke it and show you how exactly
it works. So let's take a moment till
it's done and uh then I'll quickly show
you. See now it's going to create a
brand voice reference file as well cuz
we have given the brand voice here to be
aspirational and premium. So it's going
to say everything of that sort and if
you see it's actually picking from the
from the campaign that I had put because
that campaign was about Diwali. So if
you see Diwali ready bridal season of
celebration the night before everything
it's also using like festival and
occasional language if festivities are
there right so the campaign positioning
skin that celebrates extends across all
festivals and occasion this is not a
Diwali lens so it's giving an entire
thing now it's a 500 line so it's going
to do a test prompt and it's given me a
test prompt as well check the line count
of the skill which is fine and now it's
asking me to simulate something and it's
given
an entire thing. So it's doing it. It's
packaged the skill and it's given it to
me. So now the only thing that's left to
do for me is to save this skill. It's
also telling me here's what all there
is. And let's say I go save the skill.
So now this skill is saved. And if you
go to my manage skills, I have a lot of
skills uh from what I actually do. But
here you can see Orenda real script here
is here. The entire skill lives here.
And you can see revisions whatever you
know different things of collecting
inputs is here. So now let's quickly go
back to the panel and see how we can
invoke the skill and utilize it.
All right folks, so let's quickly invoke
the skill. If I put a slash, you see all
my skills here, but the one I'm trying
to show you is the Orenda real script.
And now I'm just going to invoke it. I'm
not even going to give it any context.
Just call on that skill. And now it's
going to read the real script skill and
it's going to ask me some questions. So,
so let me collect inputs as specified.
So, it's now going to ask me questions
and it's going to take inputs. So, here
we go. Let's say what's the idea or
topic for this script. Um, let's see. I
want to go for a daily skincare ritual.
And maybe this is, you know, here is the
point where I can tell the product. So,
let's say I want to keep it general,
right? The hook should be a relatable
scenario. And let's see what it writes
for us. Most probably is going to write
something that is going to be pretty
good. So, here are the two scripts for
the YouTube shorts. That's going to be
30 seconds.
Some mornings you rush through it,
cleanser, something something out the
door. But a ritual, even five minutes,
works differently when the ingredients
are doing what they're supposed to.
Saffron and rose do not need 12 steps.
They need consistency. Your skin
responds to what you give it every
single day, not just a right before
something matters. So build yours.
Explore the full range at the link in
bio. Right. The next one is a 15-second
2 minutes every morning. The same ones
every time. That is what your skin
remembers not the right not not the one
night you went all out the ritual save
this for the morning you decide to start
save right so that's the kind of CTA now
let's say I
want to tweak this to be a little bit
more better how would I do that so on
GitHub exists a humanizer skill that's
what I'm going to use and something that
I referred in my last video. This is
Blader's humanizer. And I'm going to say
humanize
these
so that they feel like
it's butter
flowing through.
So what does butter flowing through here
mean? Basically, I want it to glide
smoothly the way that I'm speaking. So
that's why butterflowing um this is uh
sometimes you write things like this so
that it appears more conversational. So
all right hook most B most most mornings
it's barely a routine cleanser and
whatever close it's an out the thing
about saffron and rose is they don't ask
for more time they ask for the same time
every day 5 minutes done with the
intention beats 30 done in a hurry once
a week your skin is quieter than you
think it notices the pattern before you
do so it's actually telling me what it
is and now it's given me the script
right they need At the same time 5
minutes done daily does more than 30
months. These ingredients work through
repetition not the intensity your skin
catches. So now you have you see the
natural breathing area and the breathing
space. So now you understand how you can
generate read scripts. This is again
I've done it very very I would say
watered down version of it. The one that
we actually use at Upskll Nexus is my
real script generator. This is very
extensive. So I'm going to show you a
little bit about how extensive you can
develop this to be. So now this is going
to ask me the domains that I want the
information that I require that it
requires to generate a skill. And if you
see it has six kind of questions.
that I have input. So you can make it as
extensive as you would like. Now this
was the real script generator. The next
one that we're going to cover is the
caption generator. So let me quickly
walk through that one as well.
All right folks. So the next one is
going to be a caption generator. So now
what we're going to do is that we had
created our real script right um the
real script that we created was very
dumbed down. So I will create or utilize
a real script that I already have
generated. So let me just quickly go to
Orena's real script generator and say
write me a 100word
real script on turmeric and its uses for
skin care.
Right? So now I'm going to get a real
script out of it and this is the
workflow that I want to use. So let's
say I have my real script at hand.
Right? Then I want my skill that I'm
going to develop for the caption
generator to take this real script and
write captions based on the real script
that is provided. So the first step is
writing the real script uh or generating
the real script in itself. So let's
quickly do that. And um
now they're going to quickly do it. So
it's going to ask me some questions
about my hook style. and etc. So
let's say I want to say something like 2
days before a wedding and your skin
looks exhausted and this is going to
give me a 100 word one. So this is again
72 words 34 words. So if let's say uh
let's just for the sake of time let's
quickly just utilize this one. I am not
going to show you how to basically fix
this. I just showed you a workflow and I
want you to implement this for your
particular
let's say use case. So now I'm going to
write all right so I'm going to dictate
based on a real script that is provided
to you. I want you to create
captions for LinkedIn X,
Instagram, and YouTube. Paste
on Orenda Skinare.
All right. So, sometimes it does not get
the accent, which is fine. So, I'm just
going to quickly correct some of this
script that is provided to you. So I
want you to create captions for now I am
going to show you how this will do it
uh when it does not have see right now I
already have a caption generator skill
right so it's actually just invoking
that skill and utilizing it so I have
already specified so I can get straight
into the generation now I had this
already inbuilt but I am going to now
show you the process of building this
skill in itself. Right? So let's get on
with it.
Okay, let's write help me create a
caption
generator skill
that
takes in
a script
and
makes
captions
out of it. Here
is your test case. So, I'm going to give
it the test case
and
platforms
X, LinkedIn,
and let's say YouTube
and Instagram.
Okay. So, now I'm going to quickly wait.
It's going to read and it's going to
generate one, right? It's going to maybe
ask me a couple of questions the same
way that it was before. And because I
already have an existing caption
generator.
So now it's going to evaluate it, right?
And going to write one. So the core
message extractions, it's done. And now
it's written one based on the one that I
already have. So what does this tell
you? If you already have a caption
generator skill like I do, it is just
going to refer to that. Again, I
deliberately decided to show you whether
or not you require a skill every time.
But this is what it says. Now, a few
honest notes on the skill itself before
you decide whether to update it or not.
The existing skill is solid, right? But
the brand voice is generic because it
actually works for upscale nexus and not
for Oenda. Right? The ask platform step
conflicts with your workflow. And there
is no guidance or CTA. Right? So all of
this improvements. No, I do not want to
improve this. So I will just say I am
happy with the current output or I can
just leave it at that. Right? Because I
don't want to actually update it. But
you see that you can create a caption
generator skill like this. But you know
what's an additional thing that I did
not initially think of that I would like
to cover in this video is actually
turning a real script into a carousole.
Now there might be times that you have a
lot of real scripts lying in right and
you want to turn it into a carousole. So
how do you do it? So I'm going to
quickly write help me create a skill
that takes a real script
in and converts it
into a carousel.
So now this is what I'm going to do.
Let's see how this is going to be. So
now it's going to read the skill creator
skill and mayhaps
take
everything that is given to me and it's
going to also read my carousel content
generator skill. So I have two skills uh
one for the real script and the other
for carousel. So it's comparing both of
them, right? And it's also seeing that
what this particular skill needs to do
that's going to be distinct from my
existing carousle content creator skill.
Now you see the way that I have decided
to structure my entire social media
content aspect is to do with the
modality that I'm trying to you know um
use to put my ideas in front. So you
have to think of it in that way and
have that workflow in your head and how
you're going to make it. So let's
quickly just pause and take a moment
till it's generating and then we'll get
right back.
All right folks. So now we have the
entire real to carol skill MD broken
down. So what it does is it takes the
core insight uh from let's say the real
script that is linear into a kazol which
is a self-contained unit. The skill just
doesn't copy paste. It actually does
structural decomposition. It breaks
the script into parts and then maps
those slight boundaries. Okay. So now
I'm going to save the script cuz this is
a skill that I can also utilize. So how
about this? Let's actually invoke this
and try this out with the brief that we
have. So this is the thing and let's say
real to carousol enter. So now I'm going
to quickly see what's going to happen
and most probably the real to carazol
skill is going to actually build the
carousol. Now I did not specify the
output in which uh the carousol needs to
come out but usually the carousol comes
out in the format
of either content. So what it's done is
it's actually taken the content and it's
given me what kind of hues my designer
requires to take into consideration and
what kind of things I need to give to
Gemini to create using nano banana model
images. So it's given me prompts for
both and that's brilliant. One I like
for LinkedIn because the LinkedIn
version is actually in the form of a
PDF. On the other hand, the Instagram
version actually is for uh let's say in
the form of an HTML or in the form of
PNG images. Now it gives each and
everything right. So if you see this is
what it is and once so one thing I like
to do is find the factual claims cuz
sometimes when you're claiming something
AI can hinate, right? If you see here
it's written AI can make mistakes.
Please double check responses. So what
does that even mean? It basically means
that Claude can provide misleading or
incorrect responses sometimes. So you
should always double check it. And the
way I do that is that I tell Claude in
itself to fact check whatever it says by
pasting the final version back and it
fact checks each and everything that
it's saying. All right. So this is how
you can create a carousole a real to
carousole skill and a carousole with
existing content. Now this thing is not
just for converting reals to carousols.
You can actually do carousol to real.
You can do full-fledged blogs to podcast
scripts to whatever content is in your
head and whatever content you would like
to translate. Okay, before I get on with
it, I want to show you a little tool
which is called Whisper Flow. So,
instead of uh typing the entire prompt,
I'm going to show you how you can
dictate the entire prompt. Right? So, we
have seen this skill the PowerPoint one
in episode one with the campaign kickoff
tag. This time it's a debrief. So, I'm
going to take the same brand that is
Oenda and I'm going to create a debrief
tech with uh different structure from
the kickoff, right? It has leads with
results. uh moves through what worked
and what didn't and it closes with
recommendations. So, let me show you
what the re brief actually looks like.
So, I'm going to quickly come here and
I'm going to I can either click this or
I can hold function. I'm going to click
this quickly and then I'm going to
dictate it. Right? So, build a campaign
brief uh a campaign debrief on Orenda
Skincare. It's the Diwali 2025 campaign
debrief. The audience is the internal
leadership team. The campaign period was
between 15th of October to 15th of
November 2025. The results are that the
revenue was up by 41% in comparison to
last year's October 2024. That means it
beat the target of 35%. New customer
acquisition was around let's say 2,340
which beat the target of 2,000. The ro
was about 4.6 times against the target
of 4.2
and email open rate was about 31%
against the target of 28. What worked in
this campaign was Instagram reels
outperformed
static posts by three times on reach.
Influencer partnerships drove about 33%
of the new customer acquisitions. Um the
gift set messaging resonated strongly
with the gifting buyer segment. What did
not work? Well, Google display ads
underdelivered. So the spend of about
18% of the total budget uh for 9% of the
revenue was utilized for Google display
ads. The email list fatigue was visible
by week three. So open rates had dropped
by six points on the third week and the
recommendations from my side is that
increase the real production budget for
the next campaign, reduce the display
ads uh allocation to about 10% and
redirect it to the influencers.
Also set a cap on the email sequence at
about two cents per campaign and format
this entire PPT in a 10 slide debrief
deck. So show me the results first, then
the channel breakdown, then what exactly
worked, what did not, and then tell it
the recommendations and a final summary
slide.
All right. So, Whisper Flow is going to
do its magic and whatever I just
dictated, it's going to prop up on my
screen. There we go. Everything right.
So, I'm going to change this to either
highq or sonnet. Let's do highq only. I
do not want to, you know, uh unnecessary
do it. So, see claude wants to you use
generate from gamma. Now this is
something uh generate from gamma that
it's using because I have connected my
gamma using a connector. So let's say I
deny this for now. Okay. So this is not
I don't want to use gamma.
Let me see how quickly it gets done. All
right.
Okay. It's going to pick up gamma
itself.
GMA might not be available in your Let
me create this as a polished PowerPoint
deck. Okay, it's working now folks. Uh
there was a little bit of a problem with
Garma, but GMA is another PowerPoint
tool that you can use to create your
campaign decks or to create your
presentations.
So let's see. See the campaign debrief
deck has one particular job. It just
needs to tell the leadership what
happened, why it happened, and what to
do differently in a format they can read
in under 10 minutes. And this skill is
basically turning it into a deck.
Okay. The skill's job is basically to
turn this into a particular format where
they're making an entire polished 10
slide campaign brief using PPTx
genjs which is PPTX generation
JavaScript. Now it's doing it. It's
creating everything. It's removing the
buggy file. It's creating the entire
deck. While it's doing it, um I also
want to quickly address that all of this
has been covered in a previous video. So
if you have not seen it before, well,
I'll place it in the icard above. So
quickly go see the first two videos so
that you can understand this better,
especially the PBTX skill. Now the
PowerPoint presentation is created. I'm
going to quickly refresh this so that I
can see it. Uh sometimes it's a bit
slow. Why is this blank?
Let me see.
There you go. It takes a little bit of
time. So I used highq. So the formatting
is a little bit off. But if I turn this
to sonnet, it'll be much better. Uh, in
my personal experience, set does do way
way way better when it comes to creating
your PBTs. I just used HighQ because I
wanted to show you how to do this. The
Sonnet version has been covered in the
previous video. So, I would say that you
should go and check them out. Well, see
this is a tense like debrief deck and we
just saw this. The structure is very
logical is the exact structure that I
had mentioned here. So it goes from
title to key results to revenue
performance so on and so forth. Right?
But let's see now how to use the Excel
skill because this we have seen before.
Right? What we need to now venture out
into is the new skill. So let's get on
with that one.
Right folks let's see how the Excel one
works. So before I do that number one
I'm going to change this to set because
this uh is going to be analysisdriven
work. So before we get on to this right
if I want data to analyze I need to
first start by getting the data in the
form of a sample marketing data for me.
Now uh for the sake of brevity I had
already created this but I'm also going
to give you a prompt of how you can do
this yourself. So I'm going to quickly
record. So Claude allows you to record
uh your voice here as well to dictate.
However, I prefer whisper flow because
it also is an AIdriven
uh I would say dictating tool that edits
out the ums and okay. So I'm going to
press function this time and I'm going
to show you how it's done.
All righty. So,
generate sample marketing data for me.
Uh, it should have two tables. First,
campaign performance with columns for
month channel impressions, clicks,
spend, and conversions. Make it 6 months
across four different channels. Paid
search, paid social, display, and email.
Second table should be a lead funnel
data with columns for month, source,
leads generated, MQLs,
SQLs, and closed deals. Make it
realistic but messy. Some inconsistent
formatting, a few missing values, mixed
capitalization. I want to clean it up.
All right. So, here you can see this,
right? So it's uh written should have
two tables. You know how it's formatted
this. So Claude won't be able to do
that. So which is exactly why I like it.
So now I'm going to quickly take a pause
here and I'm you see it's reading the
Excel skill to create the marketing data
Excel file as well. This can happen and
I had covered this in my previous videos
where I showed you how to do this
exactly by extracting data from reports
which are PDF documents into Excel
files. Right? And let me uh quickly take
a cut here and then you can see
how it will come out. So let's wait for
this to get processed.
All right, folks. It ran the command and
it's given this um data called marketing
data. It has the campaign performance as
well as the leadfunnel raw data. The
formatting is a bit off. So now what I
want to do is I'm going to quickly
download this and save this in my
downloads as marketing data. And now I'm
going to show you how to clean this up.
Right? So let's now clean this data, fix
the inconsistencies and flag missing
values. Okay, for that I'm going to
quickly connect this right here. I'm
going to add the file and I'm going to
show you how to do this exactly.
Clean this data, fix the
inconsistencies, fill or flag the
missing values, standardize the
formatting, and build me a professional
Excel file with two sheets, one for
campaign performance and one for
leadfunnel. Use live Excel formulas and
not hard-coded values. It's very
important that you use live Excel
formula so that whenever you change the
uh specific aspect specific numbers in
your sheet it does not generate
something that's different from earlier
ones right um so I'm going to quickly
wait for this so they if you see it's uh
you can't see it here but the Excel file
I've attached it you had a glimpse of it
for now it's reading the XLS skill this
is the skill right I last time explained
what all of this is right and if you see
this is a proprietary skill which
basically means that enthropic owns this
skill so I'm going to quickly wait for
this to run and then I will show you the
difference between the truth.
All right, folks. It is here. If you see
the campaign performance completely
formatted, we saw in the last one how it
was different and the lead funnel also
completely formatted with the missing
numbers reported in terms of the
impressions, in terms of the CTR. If
there are any missing numbers, they're
completely here. And the entire channel
summary for each of them is here with
the CTR, CBR, CPC, right? All of the
important ones that you require. This is
the campaign performance and the lead
funnel. It is it has calculated the
leads that were marked in yellow. These
were the ones that were not there. If
you see see the yellow call yellow cells
are estimated missing values. So it has
done that and if you see the formulas
are all dynamic. So it's right here. And
now that I have done this I have my
clean data. Now in this right I have
shown you the difference. Let's do the
analysis aspect in this as well. So
let me quickly say uh what the analysis
aspect is and let's get on with it. So
the analysis aspect right okay I'm going
to deletate this
add a third sheet called analysis
calculate the following metrics from the
data CTR by channel cost per click cost
per lead MQL to SQL conversion rate SQL
to close rate and overall funnel
conversion rate color code the
performance green for quartile amber for
mid red for bottom and also include a
summary row.
Right, the lead funnel already has all
of this, but let's see. Um, it's going
to create a third analysis sheet. And
well, I'm going to take a step back.
It's going to inspect the workbook
structure first. And now it's going to
build the analysis sheet. It's running
the command. And it should be here in
the next four to five
seconds actually. Um four to five
minutes. Sorry, not second.
So I'm going to quickly wait for this to
run and I will see you again.
Okay folks, I actually want to bring
your attention to something here. It
says that it's computed 330 formulas
with zero errors. Now, if you see here,
it actually also recalculates and checks
for errors. There's an entire script
that sees all of that and it also fixes
the path, runs it over, it creates the
entire thing using scripts. Now, what's
a script? Well, a script is nothing but
a set of code that the machine runs in
order to execute let's say XYZ step for
you within the skill. Now if you see the
analysis section has all four of them
they're live linked. Live linked
basically means if you change numbers in
a part it automatically gets changed in
part B. And all of this is here. See one
structural note per flagging display
doesn't map into leadfunnel data at all.
So it's CPL and funnel metrics are not
applicable. If you ever get display
attribution those cells will need
formula. So see it's giving you a
warning also if something is missing
right and that's the part. So if I go to
analysis here, oops. If I go to analysis
here, you can completely see the
dashboard right of
the entire thing including the
scorecard,
right? It's all here. It's all
available. You can completely see it.
And you've seen the difference from the
original one where we only had campaign
performance and leadfunnel to now also
having analysis. Well, now let me
quickly show you a little thing that I
wanted to actually add here and this is
my little surprise in the end. So, um,
now I know what happened over the last 6
months, right? But the question is, what
should I do differently and how should I
do differently if I wanted to impress,
you know, my higherups? Here's how I
would do this. All right. So, I'm going
to quickly show you a little trick in my
book that I use to impress my higherups
and the officials, which is to build an
interactive HTML dashboard. All right.
So,
uh using this campaign and funnel data,
build me an interactive HTML dashboard.
I want sliders or input fields for
monthly ad spend per channel, target
CTR, and target funnel conversion rate.
As I adjust them, I want to see
projected impressions,
leads, MQLs, SQLs, closed deals,
projected revenue. Assume the average
deal size is of about 50,000 and blended
ros
all updating in real time and make it
clean and professional.
Okay, now I've given this and let's see
it's going to build me an HTML
interactive dashboard. So I'm going to
quickly take a pause here. Give it a
moment to generate it. Um, it does take
a little bit longer than usual and I
will come back to you.
All right, folks. It's done it. So,
based on what I just asked it, it's done
this entire dashboard, right? Um, you
might be confused that oh, this is
inside the claw interface. That's not
true. You can actually download the file
from here and it gives you a marketing
performance dashboard HTML but it does
show you the output sometimes in here.
So here's how things change here, right?
So let's say you see the sliders. Let's
say I increase my paid ad spend to about
19,000.
Reduce my paid social. Make display ads
just 1,000 emails to be 1,000. So you
see with how as you interact the numbers
below are changing right and let's say
if my target CTR then uh once I change
this right my lead to MQL rate how
that's going to be there my MQL to SQL
rate how that's going to fluctuate and
the SQL to close rate right and what my
total spend was it also gives me the
lead funnels and so on and so forth If I
do change these numbers and the
projected revenue, it says where it's
coming from. So right now I had uh put
the others as zero. But the moment like
I add this here and I change it to a
little bit higher based on my older
projections. You can see how this
changes right where the projected
revenue is going to come from based on
our past data and is doing the entire
thing. Now
my question is that what if I tell it
that based on what
I you know just modeled if I wanted some
recommendations
right um let me ask it for
recommendations give me three budget
reallocation
recommendations ations
uh let's say recommendations
in plain English right I don't want all
these buzzy fancy words so let's see if
it can avoid the buzzy fancy words and
then see this right looking at the
current default state of the model it's
xyz xyz here are the three
recommendations move budget out of
display into paid search all right email
is massive massively underinvested
relative to its output. So, invest more
in emails and don't touch paid social
until you fix your SQL to close date.
So, you see from what I just modeled,
it's given me three insights that I
should focus on. And this is uh an added
little spice that I wanted to show you.
And now, let's get on with just wrapping
this up. All right, folks. So, let me
show you the report first. This is what
a campaign report would look like. Let
me go to continuous scroll so it's
easier for you to see, right? So, it's
prepared by Meridian Media. And this is
what you receive right in your inbox.
Probably something like this. Now, this
is a dummy report obviously, but you can
see how these numbers are added in here.
uh what exactly is the performance
analysis? What are the key observations,
the strategic recommendations
and the conclusion, right? So you have
this and the table on page three here is
what you actually require. So this part
the entire table here, this is the one
that you actually require. So let me
show you how to do this and how to do
this fast.
Okay, this is my clawed interface. Uh, I
think you are already aware of what the
interface actually looks like. Now, this
is I've attached the PDF here from add
files or photos. And now I'm going to
quickly use the PDF reading skill. This
is an inbuilt skill, right? uh is that
it's going to read the entire skill and
will uh
read it, identify what's the context,
what's the data, we'll pull out the
structured information. The difference
is that it's not going to get distracted
by the agency's framing and it'll find
the numbers. Okay. So, let me uh write a
quick one right here. So what I am going
to do is that I am going to ask it
to give me the data from here. Okay. So
PDF reading skill please
extract all the important
information
from here. Now I'm using highQ for this.
Uh so you know it's checking the
uploaded files. It's going to extract
all the critical data from the PDF. So
it's already see all of this is
extracted for you. You don't have to do
much work. Now this might seem fine
right um you are extracting the data.
This is all right right it checked the
uploaded files and then it extracted all
the critical data. Now the real question
is that you want this right extracted
data and use this output to feed it
directly to the Excel skill. Now it's
the first time uh in this series that
you're probably going to see one skills
output become another skills direct
output. So keep watching this.
All right folks. So what we're going to
do is that we're going to utilize this
data that I've gotten here the
extraction into
another skill. So I'm going to use it as
an input. But because I have to change
my model to sonnet, um sonnet is good
for analysis. So I'm going to quickly
change it to sonnet. And I'm going to
dictate my prompt. This time I'm going
to use the dictate function here.
Build a twotab spreadsheet. Tab one
should have the channel performance data
extracted from this report formatted as
a clean table with all columns uh
written. Tab two
should be a channel by channel analysis
for each channel one line on whether
it overperformed or underperformed.
what the key number is and a verdict of
scale, maintain or cut.
All right.
So before I get that started, uh I
forgot to add the thing I copied. So
what I had copied here and now I'm going
to enter. So it's going to use this.
It's opened the inbuilt skill as you can
see here right now. It's going to
make the spreadsheet. So, see it's
saying it's creating the spreadsheet.
See, uh the second tab is actually what
I want you to watch this for. So not the
reorganized data because that's very
easy to do these days but actually
positioning yourself on each channel and
understanding with the numbers that
support it and how to
see that where you stand on each channel
where you're spending money. So whether
you want to scale, whether you maintain
or whether you want to cut the money. So
that's the output that probably a
performance marketer um needs at the
start of a client conversation. So it's
not a reformatted PDF but a verdict,
right? All right. So we're seeing here
it's recalculating the formulas and it's
checking for errors. Then we are looking
at this and there we go. So it is dark
themed. It is made in that manner. It
has tab one
as you can see here and it has tab two.
So we have two tabs channel performance
and channel analysis.
Now if you see here right uh there are
some things that it has given us in
terms of tab uh tab one and tab two
which are part of the analysis. Now
Instagram if you see here the ro is
about 5.8 against a campaign target of
4.2.
Now when you have this right that's
available to you. So you see the
campaign target here you see the ROAS
here. So the verdict eventually becomes
after analyzing that you should scale
Instagram because it has overperformed.
At the same time, Google display ads
underperformed, which means you should
cut back on your money in that
particular direction. Similarly, if you
see influencers, it's showing that they
have overperformed. So, they have hit
more than the target. So, you should
scale it. And what you should focus on
the mid-tier targets. So, the mid tier
mid-tier creators is somebody that you
should be focusing on. Similarly the
channel performance if you see here it's
given which all has been more uh in
terms of the returns that we have gotten
and what has been below the campaign
average. All right. So this is a
document that actually I want you to
open whenever you are going in for the
client call that is the second tab and
not the first one because this gives you
a structured positioning and explaining
what exactly it is and how uh you have
spent the money to get
the performance out of it.
Tab one, the raw data layer. Right? So
all the five channels, every column from
the original report into a clean table.
This is where you run your own
calculations ahead. If you wanted to
calculate more numbers, you can do it
here. You can even build a chart. So let
me quickly ask
uh Claude to make some charts
that are suitable
whoops
suitable
for
tab one. Okay. So now it's going to
quickly add some charts that do in the
right format. So whatever suitable
charts are there is going to do it. So
you see that it's reading the skill for
chart creation guidance as well. So it's
doing this right here. And uh after this
it's going to just go from uploading the
PDF into making this two tab
spreadsheet. Now I can give you a quick
hack as well. Once it's made the charts,
what I'm going to do is I'm going to
actually invoke the skill creator and
ask it to make a skill for the same
thing that it has done. That is take the
output of a PDF and convert it into a
channel performance report in the form
of an Excel. Right? So, let me quickly
do this. So,
build a skill that does this every time.
Reads a
what do you say? let's say reads a
report from an agency
and then makes
an Excel
with
numbers
to analyze
performance.
Right now you see it's going to create a
specific skill just for this. So if you
see here uh this is the channel
performance. These are some numbers that
it has picked up and maybe if I quickly
open this here there will be some charts
that are going to be
you know uh
alongside all right there are I see some
more things. Oh here we go the charts.
So there are four charts 1 2 3 and four
that are here and you can see. So one is
the campaign average the ROAS by channel
the budget share the cost acquisition
and the budget allocation. So you see
email had the least while the others had
the most. Now you have the data and the
charts that are available to you and the
skill creator skill right now is
understanding the workflow. So it's
going to skip the sub agents baselines
and the review and it'll basically make
a skill from here. So now you see you
can actually see here it's reading the
marketing agency campaign report. It has
made a skill. So it's reading the skill
first. Then it's extracting the data.
Then it's making a workbook out of it.
And there you go. Now you can save the
skill. It'll go right in there into your
skills section. That's you will be able
to find this in customize here where you
have all the skills residing. Uh
similarly if you are not able to find
this well you can check this out from
here and you can add it here as well.
Okay. Another way to save the skill is
to click on download and it'll save as a
dot skill file and then you can send it
to your team as well so that he can also
utilize this skill. Now an honest note
here before we move on. The PDF reading
works best on native PDFs. So documents
exported digitally from let's say report
builder design tool. Most agency reports
from wellestablished agencies are native
PDFs. So in practice this covers the
majority of what you receive. If you
want to try this on a scanned PDF or a
report that's mostly images of charts.
So the extraction will be rest reliable.
So this is something that you should
know and it is worth knowing before we
run it on something that matter.
Hi folks, we are now going to begin and
we're going to create the document for
the B2B management consulting firms. So
Kova Consulting is a B2B management
consulting firm with about 40
consultants. It's extremely fast
growing. The firm has a voice, right? It
just lives in people's heads right now
and comes out differently depending on
who's writing a proposal. The document
I'm building is called a tone of voice
guide. Something every consultant at
Kova can use when they're writing client
communications, proposals, LinkedIn
posts, anything that goes out under the
firm's name. So you see the claude
interface. you're very very uh
acquainted with this and I hope you are.
Now let me give it a quick prompt. So I
am using a tool called whisper flow
here. It lets you think of ideas, right?
It has two options of whether to polish
what you're saying or prompt engineer.
I'm choosing prompt engineer right now
and I'm going to quickly uh utilize the
skill so that
I can dictate what exactly I want.
Tone of voice document for Kova
Consulting. KA Consulting is a B2B
management consulting firm with about 40
consultants. The document audience is
all consultants plus the marketing and
business development team.
uh four tone of voice principles need to
be followed. Number one, direct and
precise. No jargon, no filler. Say
exactly what you mean in as few words as
it takes. Include a do and do not
example for that. Number two, uh make it
confidence without arrogance.
authoritative but never dismissive of
the client's existing thinking. Include
a do not and do example in that as well.
Number three, human despite the subject
matter. Consulting
can feel cold and abstract. KOA's
writing should feel like advice from a
smart colleague, not a report.
um from a firm include a do not and do
example in this as well. Number four,
practical over theoretical.
So every recommendation comes with a
concrete next step. No principle without
application. Include a do do not example
in this as well. Document structure um
should be overview section explaining
why tone of voice matters for the firm.
Then the four principles each with a do
do not example formatted as a two column
table. Then a writing checklist of 10
practical rules. Then a onepage quick
reference summary at the end. The format
should be properly structured word
document with heading hierarchy. Page
breaks between major sections. do do not
examples as two column tables with do
don't as column headers.
So it is going to process and if you see
it has everything here right it's almost
like a prompt.
So I have this here. So
now I'm going to add something more
using this document.
create a custom skill.
So let's do this step later once it's
given me the document. Right now my
focus is building the document. See a
tone of voice document has one job,
right? It needs to be clear enough that
someone who has never thought about
brand voice can read it and immediately
write better. It needs to be more than
be authentic or be like the brand
because that's not guidance or do do not
example showing the difference between a
w opening and a direct one that would be
better guidance right so watch for that
output uh we will look at that for the
output once it's done so it is creating
the document at the moment so just let's
quickly take a break and come back once
it's generated
All right. If you see here, I'm a bit
confused cuz the first page seems to be
upside down somehow. Okay. This can
happen with claw. So, I'm going to
quickly write this. The first page is
upside down.
Fix it. This much is enough for claw to
fix it. But you see um we have an
overview. I don't know if you can read
it but then we have the principle the
first principle and then what's going to
happen the dos and the don'ts right? So
similarly here principle number two,
principle number three and principle
number four right and then the writing
checklist a quick reference and what to
remove what not to remove.
Okay so
there we go it's here. So this gives
completely what it is and this is the
brief document that you can save and
then you can share it with your team. At
the same time you can also open this in
drive. So it automatically saves to
drive and you can utilize it there as
well. Well this was all for the first
thing. Now I will show you a quick trick
that you can use and you can share this
as not just a document but I'm going to
actually quickly
save this
and
all right hi friends. So what I'm going
to do now is that I am going to utilize
the dox skill and create the other three
documents. the brand voice system, the
brand kit foundation and the brand audit
proposal the all three of them so that
we will understand how we can utilize a
docx skill to create it and at the end
using these documents that we have
created we will create a custom skill
that is going to be helpful for my
entire team which has access to claude
to automatically format at and adhere to
these guidelines. So, not only are we
going to give them the guidelines, we're
also going to teach them how to utilize
it. Now, um the second brief is a little
bit more demanding. Uh so, the brief
that I'm going to give Claude, it's
going to ask Claude to make judgment
calls about what KA should and shouldn't
sound like based on the company's
positioning and the uh and the
competitive context. So what I have done
is I have written the entire prompt for
you already. Uh I am just going to
quickly copy paste it here. And if you
see here the brand voice system for KA
consulting, it's given the company
positioning. So what exactly does it do?
What it is not? What's the comparative
context? What's the document structure
going to be? So there's going to be a
vocabulary guide, a sentence structure
guide, a register by channel because
what you write on LinkedIn cannot work
on Instagram and what you write on
Instagram will not work on LinkedIn. And
then section number four. So what good
co writing would sound like. I'm going
to quickly give this and wait for it to
run. While that is running, what I can
do is I can actually place all of these
conversations into a project. Now,
what's a project? A project is a space
where I can keep all my conversations
that are related to one particular topic
in a folder. So now you see it says KA
consulting/brand voice system. I'm going
to do the same thing and I'm going to
add this as well into the KA consulting
project. So now when I go to project and
I go to Kova Consulting, you see that
you will be able to see this. Now this
says that it wasn't sent. It's because I
moved it.
Oh, it is working already.
Is it? It is working. Just give me a
second. I'm going to quickly send it.
And now it's making the entire brand
voice document. Okay, while that is
happening, I'm also going to open up
another new chat and I'm going to give
it the prompt for the brand kit
foundation
and the prompt for the brand audit
proposal. Now mind you, the brand audit
proposal is actually a proposal that is
for investment.
So uh the situation usually in a brand
or a proposal will be related to a PDF
and not a DOCX. So I'm going to wait for
that one because these three are the
only ones that we're going to utilize to
actually get a skill made for the brand
voice.
Okay, here's how you're going to go
about creating a skill using all the
docs that you just created. So, I'm
going to attach the four
the three documents. I keep saying four
cuz I'm teaching you four skills, but
the three documents. And I'm going to
invoke the skill creator and say create
a specific skill
that captures
all the
aspects
in these documents. Uh so instead of
attaching these documents here you can
also attach them here inside the project
so that every time you invoke it invoke
something let's say a conversation so
instead of attaching it here right you
can directly add this
in while you are reading this out. So it
will now reference those three documents
to
understand how to create the skill. It
may ask you some more questions. So let
it ask the questions and then it'll
basically talk about how it is right. So
it's now building a tight well
ststructured skill that captures all of
them. So now it's going to use the main
skill MD file. Okay, this is the skill
MD file and it's going to write
the main skill MD file of what it is,
what it has to do, what it doesn't have
to do, etc. And then it's going to give
you all the information. So, it's read
the skill creator instructions. This is
an inbuilt skill in Claude, right? And
yeah, so now you basically wait for the
magic to happen. Once the magic happens,
you have this ready for you to go. Now
while this is happening, right, uh it
can run parallel things. So now what I'm
going to do is I am going to give it a
document where I want actually a PDF to
be created. Right? Now this we were
working with dog access. Now let's talk
about a PDF and how you can do that. So
let's say the situation right so you're
Kovas brand manager and they need to
make the case to let's say six senior
partners that the firm should invest in
formalizing its brand the partners are
time poor and skeptical of the brand
spend right so Google doc is not going
to do the job for you you will have to
make a proper PDF report note what this
document actually is is actually a brand
investment proposal and not a full audit
a proper audit is about 20 pages of
research findings Right? This is a tight
executive facing document that makes the
case for commissioning one. Call it what
it is when you're briefing and the skill
will produce the right document. So I
had already generated
uh using my prompt engineering bit
a complete
briefer of what my investment proposal
should look like. And I'm just going to
quickly send it and it's going to do an
entire audit and going to use the skill
MD for PDF creation and create a PDF.
See the brief is written the way you
would brief a consultant. So there are
specific findings, the specific data
points, explicit audience that is
mentioned, explicit purpose that's
mentioned, the 71%
uh client survey that is mentioned there
as well. And um
you know it's going to be the stat
that's mentioned is going to be very
interesting for the senior stakeholders.
So there is specific numbers that make
the case in a way that the vague
observations don't. The stat
that stat came from a fictional internal
survey for KA but the logic still holds
right. So find a specific number and put
it in the brief so that you can utilize
it for your brand. And now I'm going to
quickly Okay, by the time my
formalization business case is happening
here, I am going to quickly look at the
brand creator and the brand writer skill
that they have made. They have utilized
all the different things and now I have
a skill that does it for me. In fact,
your brand references, it is full of
guidelines that we had first generated
with the breeze. Right? I could have
done this without the documents also,
but the documents can act as great
references. So now what I'll show you is
how you can manage the skill and add
references here. So for example, there
is a markdown file, right? So let's say
in the skill I want to add more files.
So I'm going to say uh something like
edit with claude
and let's see what happens. I have not
tested this out myself so I am figuring
this out with you.
So, I'm going to quickly put these three
and say
I want
to add these as
reference
documents
in
the
in the skill,
right? You can we have also talked about
how you can tweak the actual skill
and let's see whether it's going to be
able to add all of these
texts from these documents into the
brand references. So while it's doing
that um
brilliant
now you can cross question it.
See?
Yes.
So now it's going to quickly just edit
it. And once this is done, now I've
shown you how to edit your skill from
the back end as well. So now it's adding
the parts that were missing and you can
see completely what's going on.
Okay, one thing that is worth flagging
here is that I have created this right
but all of these numbers are
illustrative and the ranges are also
illustrative. So it has used this PDF
using report lab and it has made the
entire PDF. So now you see that with the
briefing like this where you tell it the
key findings what it to do and they kind
of create like an executive summary
report that can be done right here right
and it's going to be extremely easy. So
it's given a proper thing for now the
formatting may be off here and there
which it is. Uh I have no idea why what
is happening with claude in terms of
it's
uh see it's now all right sometimes you
just have to refresh it. It just goes a
little haywire right but now you have
this entire document and this is
something that you can send to your
brand manager to actually uh to send to
your hire officials not your as a brand
manager to get some money that you can
utilize to make your life better. Now um
so what we've seen is from a brief to a
propos to a professional PDF proposal
and that to in about 5 minutes. So this
is what it is and this is what it will
look like across a brand manager's ear
uh you know across a brand manager's ear
of the kind of things that they're
doing. You see brand managers produce
the same document categories
um on repeat whether it's guidelines,
voice systems, briefs, reports,
proposals. It's just that the format
varies but the pattern does not. We're
going to be covering five skills today
from one source. So, I'm going to have
the agency report from the Midian Media
Orena campaign and we're going to read
the PDF, analyze it using the XLS skill,
make a leadership deck using PPTX, have
a LinkedIn post crafted, and then use
the humaniz humanizer skill by Blader
Insight Claw to make it human and
natural and ready to post. Right? So the
same campaign data with zero tool uh
switching end to end I'm going to be
covering in this video.
Okay before anything let me be specific
by what I mean by chaining because it's
more important than just using multiple
skills in a row. Right? So in a standard
workflow right you probably use a tool
you get an output. So if I were to say
here you have an agency report then
you'd probably copy paste it reformat it
and then send it to the Excel where you
will again rewrite and adapt and build
the slides in PowerPoint. Then maybe you
would use this you know to write uh your
LinkedIn post and then probably strip
the AI tone and decide to humanize it.
Right? So each tool has a gap right uh
between the next one. Then this is a
human step that prepares the output for
the next input. A scale chain
essentially eliminates these gaps.
Right? The output of each skill is
already formatted as the input for the
next skill. So you're not preparing
anything between the steps. So you're
just handing it off directly, right? So
the manual handoffs get eliminated and
the hours disappear. So now you have one
chain with no gaps and the speed
automatically compounds. See the analogy
that I keep coming back to is the
difference between a relay race and a
series of sprints. Right? In a series of
sprints, each runner finishes, hands the
bat to let's say the next to the the to
the maybe a coach and then the coach
decides what to do with it and
eventually the next runner starts.
Right? In a relay on the other hand
where it's not a series of sprints, the
first player or the first runner
automatically gives the batn in the best
way that can be carried by the second
runner. Right? The batn goes directly
from hand to hand. Nothing is lost
between the steps. So the speed
automatically compounds here. And that's
what going to what we're going to see
today. five skills. Each one's output
becomes the next one's input. So, it's
one campaign worth of data processed end
to end without you managing the handoffs
manually.
Before I even run the chain, I want to
name what this workflow looks like
manually. See, I don't want to be
dramatic about it just because the
contrast is going to be the point. Well,
you receive the agency report, right?
and it probably receive it as a PDF.
Once you open it, you'll probably read
through it, find the entire narrative,
right, with the actual numbers, the
copy, the performance table into the
Excel sheet, right? Reformat the columns
maybe to match your own naming
convention and this would take
approximately 40 minutes, right? Then
let's say you want to build a debrief
deck in PowerPoint. So, you maybe
already have a template if you, you
know, uh are somebody who has their own
brand strategy. In case you haven't seen
how to make your own brand strategies,
um well, there's a video in this same
series. I will link it up here so you
can go watch that. And uh let's say you
have your own template, right? You
create the slides from the results. You
have the channel breakdowns, what
exactly worked, what didn't, and the
recommendations,
everything, right? you write it down
that's another 90 minutes off the bat
right minimum if you're working quickly
I'm talking about this 90 minutes a
minimum sometimes uh you know my friends
in corporates they would spend about a
day just making a deck right let's say
you do this as well right and now you
think that oh I want to write a LinkedIn
post out of it um seeing what are the
key insights from let's say this
particular campaign so you draft it then
you edit it then you strip the AI tone
out of it, right? If you use something
like Claude 2 drafted and that's another
30 40 minutes spent there. So total
somewhere between 2 and 1/2 to 3 hours,
right? Spread across 1 2 3 four tools at
the moment without
uh you know using any kind of uh AI
tool. you'd probably be using multiple
tools having manual hands off and the
time gets lost in these gaps right so a
scale chain workflow a connected one
here as I have shown would probably work
better than the manual workflow
I'm going to be using the same agency
report that we used back in episode
number six you can check that out up in
the icard the episode number six if you
haven't seen it but we were working with
Meridian media's Diwali campaign rap for
Orenda skincare so if you haven't seen
that episode this is basically a
fivepage PDF with a narrative right and
a channel performance table inside it so
the five skills that I'm going to be
working with I am going to be using
claude for that so I'm going to quickly
attach this here um it says that I'm out
of usage credits but I am Not uh I
checked my settings. I think there is
some issue from Claude's end that this
error message is popping up. Um all
right so I have attached this in the
chat section right it's asking me how
can I help you right so let's uh look at
PDF reading first so I am going to make
claude read the PDF right so I've
uploaded the report and uh you've seen
this step before multiple times right so
the skill reads through the PDF pulls
out the structured data it ignores the
pen padding all of it Right. And I'm
going to ask to uh for it to make a two
tab spreadsheet. So I'm going to say
turn this PDF into a two tab
spreadsheet,
right? Uh
what all do I need to have? So tab one
is going to be let's say about channel
performance
data
from this report and tab two for that
matter let it be channel by channel
analysis right
uh I want the verdict of scale let's say
maintain
or even cut for that matter
for each channel,
let's say with the key numbers
uh that support it.
Right? I'm going to be quickly running
this. If you see that the error message
is there, but it's still working. There
is some backend problem, but it is now
working. It's reading the Excel skill
and it'll also use the PDF. Uh, so it's
already extracted it from there and now
it's going to be turning this PDF into a
two tab spreadsheet.
So also from episode 6, right? Raw data
in one tab, channel vertics in the
other. If you watch that episode, you
know exactly what's about to come. So
watch what happens after this one.
All right, good. So now I have it in the
form of let's say a spreadsheet. It has
all the channels that are mentioned, the
performance of each channel, what it is,
whether to scale, whether to cut or
review or even maintain. Right? Um, we
have the analysis and the performance
both in different tabs as requested. If
you see here, Google display consumed
about 9.1 times the return on ad spend.
So, definitely cut or review. So uh the
verdict is cut or review because it's
about you know uh consuming 18% of the
9% of the total revenue share right so
the verdict it's cut and that's the
inside and we'll hold on to this because
this is what we will need in step number
four right now let's move on to the PPTX
skill uh I have a prompt that I already
written about let's say a campaign
debrief tech for leadership team the
source data is the channel performance
uh analysis
above. So I'm just going to quickly
above. So I'll lead with the headline
result, the return on ad spend, uh
whatever there is and all of it, right?
I need 10 slides in total. So I've
mentioned that and I've given this to
them. Now here's what's new, right? In
episode 1, I wrote a brief by hand to
feed it into PBDX. In episode 5, the
marketing manager probably wrote a de
debrief brief. This time the data from
Excel analysis is the brief. So the
chain is actually feeding itself. The
output of step two became the input of
step three without me writing anything
new. Right? Uh so if you see if you if I
take a moment here it's already
generating and once it's generated I'll
be back.
Okay folks the output is here and here
is what you can see we meet the target.
So that's good. They've given the entire
campaign right in different uh methods.
They've put this here as well. So there
are three recommendations right one is
scale Instagram. Number two is uh
cutting or restructuring the Google
display ads or capping the emails to two
cents. Right? Um and also refining the
influencer creator mix here. Right? So
that's the debrief deck. So every time
that you know somebody wants to actually
make this they can directly use it. So
they have done this. Now let's
understand uh how we can use this
debrief deck as let's say uh something
that's going to allow me to attach this
debrief deck in let's say uh LinkedIn
and uh write
let's say uh LinkedIn post related to
this and how we are learning the lesson
from let's say our last campaign uh
about Google display ads and what
happened and how you know we have
figured it out. So I had made this. So
now let me just quickly put this in. So
it should be professional direct and
insight first. So it's reading my
humanizer perfect skill. So this
humanizer personal skill is basically a
skill that I created uh so that it get
the general humanizer skill also gets
refined according to my own personal
tone. So now it's going to be using
that. So using the same caption
generator skill right for LinkedIn that
you saw from episode number four. The
source material has changed but in
episode 4 we were generating let's say
from social content uh you know for
let's say a particular campaign topic.
Here what we're doing is we're
generating it from specific data point
that came out of an Excel sheet. The
insight isn't invented at all. It just
came from the chain. Right? So let's see
here. What does it say? We ran Google
display with 18% of our budget last
campaign. It delivered 9% of revenue.
ROAS was this much. Campaign average was
4.6.
Right? We're cutting it for the next
one. Uh and honestly the bigger lesson
isn't about Google display. It's about
the habit of keeping a channel in the
mix just because it's there because it's
familiar because it ran last time. So
this is writing in the same tonality
that I write my papers in. So good
structure, right? It's also insight
first and there's a specific number.
There's a clear decision. Now the last
step of humanization is already applied
in here because it automatically invoked
my personal humanizer skill. uh because
I'd already written something about
tone, right? Let's say if uh I were not
to do this, right?
If I were to, you know, not invoke let's
say the humanizer personal skill here, I
probably just say /humanizer
and write something like humanize the
above answer, right? And that's how it
would be. So now we've seen the version
that goes out that is the version that
is completely humanized
from
uh you know the Excel which was then
converted into a debrief PBD. So we have
seen five steps coming out of one agency
PDF in so the spreadsheet the leadership
deck and a LinkedIn post out right total
time for all of these five steps were
probably the time uh that was this
entire video. So here you know
we are not doing the same work that can
be done manually uh across different
tools with reformatting in the gaps
right that's the better part of the
morning and this is how you can optimize
your time and save up time in your
workday.
See, the practical implication of a
chain like this is not just speed. It's
the category of work that used to
require blocking out a morning that now
fits inside just 30 minutes. That
changes what you actually put in your
calendar. So, a campaign brief that
previously needed only 2 hours of
focused document production time can
happen the same morning the report even
arrives. And the LinkedIn post as well
can be crafted that could have been
delayed you know can be crafted
in the same time that someone had to
write it um comes out of the session as
a debrief right so one honest note on
these change uh chains though they
require clean handoffs so if your source
PDF is poorly structured and it's heavy
on images or is scanned and rather than
exported the extraction in probably step
one will be rest reliable and that will
then affect the entire chain. So it's
extremely important to also check the
output at each step and the input
automatically then gets checked right
the chain is only as good as its first
step for well structured agency reports
right which is probably most of them
this will run cleanly for many source
documents you might need to tidy the
step one output before feeding it
forward. See the broader point is that
start looking at workflows not as a
sequence of tools but as a sequence of
handoffs. So where are the gaps? Where
are you doing manual reformatting
between steps? Those are gaps that can
be fulfilled by building chains. Before
anything else, right, I want you to
first understand what actually a skill
is and what are the components that are
part of let's say this skill. Right? So
understanding the anatomy of it. So
number one is a trigger condition,
right? So when should the actually the
skill activate, right? Whether it's when
a campaign brief is input, whether a PDF
report is there or a brand review is
requested. So understanding what
activates this skill, what kind of
output uh is required from it comes at a
later stage. But understanding what kind
of input or context tells Claude in its
skill mode rather than you know uh the
general chat mode for a production skill
this is usually a file right and usually
a file type or a task type so a campaign
brief or you know a PDF report for a
review skill the kind where you have to
actually review work it is probably
content plus the instruction to evaluate
it now the second is what is the task
right because having a trigger just
means that you activating it. So imagine
it as a domino. So the first part of the
domino is like you know when you get
dominoes to fall the first step that you
do is actually take your finger point it
at the domino and then push the first
block. So the pushing the first block
the act is the trigger. The second is
how do you define a task right? So the
task of all the dominoes that are
falling one by one. So what does the
skill actually do? Not weekly but
precisely. So instead of saying, you
know, review this copy, well, that's not
a task definition, you'll probably say
something like check this copy against
the following brand voice criteria and
flag anything that viates them, right?
And that do with a specific reason and a
suggested alternative. That's a task,
right? And that's a task definition. The
precision of this component determines
whether the skill produces useful output
or generic feedback. Now the third one
in this uh you know anatomy is the
output. So what should the actual even
result look like? So the result of all
the dominoes falling is the dominoes
putting themselves in a pile. Now here
we will understand what does the result
look like? What format what sections
what level of detail you know a
production skill output uh probably is
going to be a file. A review skill
output could be a structured verdict and
that structure of that verdict also
needs to be defined otherwise you'll
just get like one big chunk with you
know a paragraph of general feedback
that's even useless. It's not even
something that you can you know act on.
And then the fourth component in this
entire anatomy is the rubric that is
what standard should it follow. Right?
So just getting it to create an output
does not necessitate that the output is
going to be correct. It could be that
the output is wrong. Right? So you need
a fourth component that basically
reviews these skills based on brand
voice, compliance rules, formatting
standards or even checklists. Right? And
this is going to act as a rubric, right?
The standard against which the content
is judged. Now this is what makes a good
review skill. Okay? and it's going to be
genuinely useful rather than asking some
sort something like okay is this good
because a rubric actually contains your
criteria like your band phrases your
tone principles your specific dos and
don'ts right without a rubric the skill
is judging against its own general sense
what is good writing so you can't really
judge whether the copy that you've
written is actually good or not but with
a rubric it's judging against yours and
that's the difference between a
pre-built skill and a custom one
pre-built skills have these generic
standards that are kind kind of baked
into the system and a custom skill has
your standards baked in. So it's very
very specific to you. Now let's quickly
understand of how we can do this using
claude. So let's get on with it.
Okay. So before I even get on with
actually understanding and developing
this custom skill with claude, if you
look here, here is you know a quick
snapshot of a difference between a
custom skill and generic AI and why
prompting generic AI can cause problems
and why a custom skill is needed. I've
already explained this in the previous
section but if you want to just quickly
go through this you can take a
screenshot even even save it for future
so that you can understand what exactly
happens here. and what happens in these
systems when you actually create a
custom skin. So this is what a dashboard
looks like, right? We've already seen
this. We have been very familiarized
with this. We also have different kinds
of models. We saw how we can utilize the
different kind of models for specific
use cases. Now uh this one I have
specifically created a this this time
I've specifically created a free uh
claude account so that the viewers who
are viewing this who may okay uh we are
in the panel and I have opened the site
pane in the side pane you can look at
customize. So when you go to customize
you see the skills that are available to
you. Now before I even click skills
right you can already see here that
Claude actually asks you how you can do
this. So you can connect your own apps
you can create new skills and you can
even browse plugins. Now what are
plugins? These are basically pre-built
knowledge for your field. So in case you
want to understand how the entire claude
ecosystem works, you can drop a comment
below and I can explain how you can
utilize Claude for your benefit and
especially for your roles. So the kind
of roles that you're interested in,
please remember to drop a comment below
so I can actually uh look at what kind
of videos you know our audience wants to
write your own set of skill
instructions. So a skill like we
understood in the previous videos is a
markdown file. So you explicitly tell
Claude in markdown what all is required
based on the um file structure that a
skill understands and then here you can
directly upload a skill. So we had
already uploaded the humanizer skill
which was written according to somebody
else's standards. So now I just quickly
click on create a new skill with claude
right and it's given this let's create a
skill together using a skill creator
skill. Well that's a mouthful first ask
me what the skill should do. I am just
going to quickly enter this and let's
see what Claude responds. So now Claude
is going to give me I'll use the skill
creator skill to guide us through this
properly and let me check it first.
Right. So now it's checking what the
actual setup is how it's going to be and
it tells us you know what we're going to
do testing etc etc if you can see you
can see here the trigger context the
output and the test cases now the input
okay the anatomy that I mentioned here
the trigger the task the output and the
rubric if you see here all of them are
here the trigger context so when should
it actually kick in what is the output
going to look like so what exactly is it
going to do and what are the test cases
so test cases are basically checking and
reviewing for whether the skill that we
have created is working in the right
direction or not. Now, uh the skill that
I'm building today is going to be a copy
review skill for Orenda. So, this is the
brand that we've been using across this
series. And the idea is that any piece
of copy going out under the Orenda brand
name, right, gets run through the skill
before it's even published. it gets
checked against ora specific brand voice
criteria and structured um basically a
structured output that is going to be
actionable. So this is the architecture
that is somewhat in my mind right that
is to build a copy review skill for
Renda right so we're going to do this
entirely and this is the architecture so
my trigger should look something like
review this for reinda is this on brand
something like that then I'm going to
look at the brand voice rubric then I'm
going to review it right and then I'm
going to give it a structured outward
now this is the architecture for the
skill that I'm going to use all right
folks so instead of actually typing the
entire prompt. I have already, you know,
written it somewhere for myself. So now
look at this. What am I saying? Build a
copy review skill for Orena. Right? The
name of that skill is going to be this.
What is the trigger? Anytime the user
paste or uploads or copy and then asks
for a review brand check or a
pre-publish check, right? Also to be
triggered on phrases like is this on
brand, review this fora or check before
I send. What it does exactly? It
basically reviews any piece of forenda
copy whether it's ad, email, social
post, product description, anything,
right? It checks against brand voice
criteria, right? So, I want it to be
more direct and warm. I wanted to be
specific about the ingredients and
benefits. I want no fear-based language
in it. So, something like problem skin,
anti-aging, damage, nothing. These are
all I would say phrases that are banned.
Then uh from I'm going to also talk
about you know whether it celebrates the
customer's existing routine rather than
positioning against the alternatives
that are there and then also I want the
sentence rhythm rhythm to be short lead
sentences with supporting details right
then the vocabulary what kind of you
know words I don't want it want it to
use words like clinically proven right
because as a skincare brand it could be
possible that it's not yet clinically
proven it's under it's under trials and
that's what happens with most skincare
brands So I need to be very careful of
what actually I put out and then I have
basically created this and then the next
thing I have you know after the review
section of what actually it's going to
review the output format I have said
that okay you know pass pass pass with
edits or review needed there are three
things right that my senior copywriter
was doing which is either they were
passing it or they were saying that it
will pass with certain edits and an
entire rewrite that could have been
needed for that particular copyright. So
you flag the line and then you basically
give the tone summary of whether even
the overall piece sounds like orinda or
what orinda is supposed to sound like.
Now I'm going to quickly hit enter on
this and we are going to now come back
and see when and you know why it while
this builds when this completely builds
out. It is also going to ask me certain
questions right uh whether it should
check to all paragraphs. I'm going to
say all paragraphs for this one. Uh I
want all variants of this. Okay. Do you
want to test cases before we finalize?
Yes, I want to also set up the test
cases. So, it's also going to ask me
prompting questions that my instructions
may not actually specify. And now it's
done everything and now it's setting it
up. Right now, you see that the skill
creator is taking this specification and
then turning it into a configured skill.
One that probably, you know, Claude can
trigger automatically with all of these
criteria already loaded. So I don't need
to paste these brand guidelines into
every review request now. They basically
live inside the skill. So I'm going to
quickly wait till this you know creates
the entire skill MD and I'm going to be
back then.
All right. Now you see that there is a
skill.md that is there right? It's
already drafted. It has done the three
test cases. It has checked it entirely.
Now what is it asking me here? Are this
thing? How do these look? any suggested
and I'll say no it's all right
build me the skill now so it's actually
not yet built build the skill it's only
build the skill.md and now it's going to
run the script and build it into the
proper directory so now it's going to
package it so once it is actually
convinced or it has convinced me in fact
that the skill is actually you know
going to be used ahead then it now Once
the validation has been passed, it now
creates a writable output location
something right. Okay, while that is
happening, let's quickly also write a
bad copy. So, uh I want you know I have
this bad copy that I had written
here. So, I'm going to quickly paste
that as well. Now, if you see it's done,
see the name is exact same orenda copy
review. All the criteria that I
mentioned uh is mentioned here. So you
have it needs to be direct and warm
specific right no broad claims etc
vocabulary to f you know to flag and
stuff like that everything is here and
this is the output format now I'm going
to save this table quickly okay and I
had already written a bad set of copy so
this has deliberate errors in this so I
want you to I want to show you how this
works so I can just invoke it by the
slash and I've copied it so you see the
word revolutionary is here which is bad
transformative. So now it's reading it,
right? And then it's reviewing and then
it's given its verdict. What is the
verdict that an entire rewrite is
needed, right? So this is going to be a
piece of the copy. And here are the
lines that are flagged. It says what?
Revolutionary banter. You need a
suggested alternative transformative
formula. Right? So it's going to say
things like signs of aging are you know
um see again signs of aging fear-based
language alternative softens fine lines.
So it's giving you how you can fix it
right. Um it's saying that the suggested
alternative also you know needs to
probably be framed around skin gains
that is not understanding what you're
fighting against but what you can gain
and that's usually a brand's you know
voice principle uh that is going to be
showing up in this rewrite. Now this is
a good catch right that I want you to
see here that it is not just looking at
single phrases it's also looking at
context. So a simple single phrase can
also like show up in two different
places. So one is let's say naturally
nourishing powerful blend of active
ingredients right here the word of
natural you have to mention it like what
exactly is there so it's like an
unspecific claim right and because there
is no active ingredient that is there
that criteria is also something that is
you know broken. So now it has dissected
it completely and this is the kind of
non-generic grammar checker thing that
we actually want this to do. It's to
review at a higher level. Now, um you
understand the tone summary, right? Uh
the two sentences on whether even it
sounds like or it says that it's
completely bad when it comes to sounding
like the brand. So, every violation gets
caught, right? Everyone with a specific
reason, a specific alternative is also
proposed for it and that's the skill
doing the job and that currently falls
probably to a senior copywriter or even
a brand manager, right? Who's reading
every piece of coffee before it even
goes up. Now, here's probably what the
corrected version looks like after
applying the flags. Now, what I'm going
to do is I have the corrected version
with me and let's quickly paste this in
and actually ask it to evaluate this.
Okay,
let's see what's the verdict on this,
right? U maybe it's going to be better.
Maybe it's going to still say see
overall verdict passed because this
actually sounds like the brand. So, what
does it have? It has the orinda glow
serum what there is. It tells you what
all ingredients are there and then it
talks about how it can fit into your
routine. So this version would
definitely pass the skill. See the copy
review skill is actually one example.
Before I give you the broader picture,
one distinction that's worth knowing is
what category it actually fits into
because it changes which kind of skill
you would build for what kind of a
problem. Now there are three categories
broadly. If you see here you have
production based, review based and
transformation based skills. Now
production skills generate any kind of
output. So whether it's a document, a
deck, a report, right? Everything that
we've covered from episodes 1 through 8
in this particular series. Then the
review skills actually evaluate. So this
is something that we have seen right now
with the copy reviewer, right? So they
flag problems, they suggest
alternatives, they give a verdict
because that's how a review skill is
actually meant to be built. And then the
last kind is the transformation one
where we see that we take an existing
output, right? And then we rework or
repurpose the content for a new
particular need. So whether it's
humanizing, whether it's changing or
creating a LinkedIn post from your
brief, right? We've saw we've seen this
in the last you know couple of episodes
and that's something what a
transformation skill does, right? So
most marketing teams, well, they at
least need one of each. So here's what
you know what's worth building in my
head. Um because you just don't want a
chatbot that's giving you answers. You
want a workflow engine, right? Somewhere
you're giving an input, then you have
the rules set in, then you review the
logic and the output format, and then
you can actually get results into your
workflow that are extremely important.
So if your brand is specific right to
let's say um is specific enough for a
particular use case which is going to be
in most cases then instead of using the
humanizer skill right you can maybe
create a brand voice skill then find a
skill that not only just removes the AI
patterns but a skill that checks for it
right and then in terms of production
you can think about campaign briefs if
your team is producing campaign briefs
regularly and they always need to follow
the same format right then a campaign
brief is something that you should think
about creating so with your section
ctions, your names, your field names,
your agency's requirement, everything
that's baked into it, right? If you work
with clients and produce reports in
let's say a specific format, right? Then
you can have a skill that basically
changes it into a specific format. So
you reformat it for the client facing
thing that you require. If you operate
in a regulated category, for example,
BFSI or healthcare or even FMCG for that
matter, right? anything that has any
kind of compliance checks. Then you can
create a compliance review, an evaluator
skill that basically evaluates all the
pre-compliance claims that you are
making and all the claims that you could
make. Now, these are probably, you know,
skills that will also help your legal
team to review these. So, you're not
essentially replacing the legal review.
You mind that. It's about catching the
obvious problems before it even reaches
legal, right? So that legal doesn't just
point out the obvious things and be like
dude this is so obvious right it's just
going to save time for them and for you.
So the principle across all of these is
just the same. So you find your task
that your team does on repeat where a
single person is currently the quality
control layer. You basically standardize
it for everybody and say okay this is
how we're going to go ahead. This is how
the skill is going to be built and this
is how you're going to use it. So
instead of creating a first pass editor,
you basically create a reviewer, right?
And this is going to be the last demo in
this series. Um where we have unpacked
all the different kinds of workflows and
we've also unpacked what exactly it is
and how exactly we can create our own
skills for our particular workflows as
well. Let me say something to you
clearly and what I think you should
actually take from this entire series.
See, I don't want you to get lost in the
volume of content. There is so much
content on the internet itself. But this
wasn't a series about prompting either.
It was not about even getting better
outputs from Claude, you know, by asking
better questions. Um, the whole argument
from episode 1 all the way to through
this one has been about a mental model
shift from using Claude as a chatbot
that drafts for you, right? and it
drafts things for you that you can work
with to using it actually as a workflow
engine that produces finished
deliverables. That shift basically
changes how much value you get from the
tool. You know, not by just a tiny
margin but by a multiple I would say
compounding kind of an effect and value
that it can generate. The one thing that
I'd ask you to do this week, right, is
to pick one workflow um that your team
does on repeat where the output is
currently inconsistent or probably takes
longer than it should, ask the three
questions. So, does it repeat? Does it
have a specific format? Does someone
review it for consistency? If the answer
is yes to all three of them, build the
skill. It'll probably just take an hour
for us, you know, if you get well versed
with it. It took about just 15 to 20
minutes to do this and the return on
that 20 minutes probably compounds every
time the workflow runs. But before you
do a quick question for the comments,
what's the first custom skill you're
going to build? Drop it below. You know,
I read every comment in this series and
I actually want to understand what all
new content you want to see on this
channel.
And now that this series is closing,
I'll just say the question isn't whether
AI is going to change how marketing
teams work because we already know it
has. The question is whether you're
using it as a chatbot or as a workflow
engine. And I will leave it with you
guys and I will see you in our next
series. Don't forget to subscribe to
this channel to get updated with the new
series that we come up with and we also
conduct lives every week. So do check
them out as well. Thank you so much and
I hope you really enjoyed the series and
I will see you the next time. Bye-bye.
Ask follow-up questions or revisit key timestamps.
The video argues that most users underutilize Claude by treating it as a chatbot rather than a powerful workflow engine. By transitioning from simple prompting to using 'skills'—preconfigured, repeatable workflows—marketing professionals can automate end-to-end tasks like creating stakeholder decks, generating social media content, and processing reports. This shift moves the focus from manual production to efficient execution, enabling users to create finished deliverables rather than just raw text.
Videos recently processed by our community