1
00:00:00,000 --> 00:00:05,248
Welcome to your first session with Claude
Code, from K and A Performance.

2
00:00:05,248 --> 00:00:10,043
By the end, you will know how to open a
project, give Claude a useful brief,

3
00:00:10,043 --> 00:00:13,224
review a change, and check the result
yourself.

4
00:00:13,224 --> 00:00:16,150
Our practice project is a small reading list.

5
00:00:16,150 --> 00:00:21,003
We will ask Claude to make its checkboxes
work and remember their state after

6
00:00:21,003 --> 00:00:26,413
a refresh. Claude Code can read files, edit
code, and run tools inside a

7
00:00:26,413 --> 00:00:30,082
project. That makes your instructions
important.

8
00:00:30,082 --> 00:00:34,749
Start with an outcome, a boundary, and a way
to verify success.

9
00:00:34,749 --> 00:00:38,882
For example: make the reading checklist
remember my choices.

10
00:00:38,882 --> 00:00:40,774
Keep the existing design.

11
00:00:40,774 --> 00:00:43,259
Check that the choices survive a refresh.

12
00:00:43,259 --> 00:00:46,927
You do not need to memorize every command
before you begin.

13
00:00:46,927 --> 00:00:50,967
You need a small task and a clear definition
of done.

14
00:00:50,967 --> 00:00:55,240
In the interactive lesson, the practice
terminal is a simulation.

15
00:00:55,240 --> 00:00:59,930
It does not run commands on your computer or
send anything to Claude.

16
00:00:59,930 --> 00:01:04,365
Use it to rehearse, then try the same
workflow in your own project.

17
00:01:04,400 --> 00:01:06,629
Let us get the terminal version ready.

18
00:01:06,629 --> 00:01:10,054
A terminal is simply an app where you type
commands.

19
00:01:10,054 --> 00:01:12,353
On Windows, open PowerShell.

20
00:01:12,353 --> 00:01:17,113
On a Mac, open Terminal. Choose your
operating system in the lesson and use

21
00:01:17,113 --> 00:01:19,818
the linked official installation
instructions.

22
00:01:19,818 --> 00:01:24,404
The Windows PowerShell installer and the Mac
or Linux shell installer are shown

23
00:01:24,404 --> 00:01:29,315
on screen. These commands download and run
the official installer, so check

24
00:01:29,315 --> 00:01:31,683
the address before you use them.

25
00:01:31,683 --> 00:01:36,896
After installation, open a new terminal and
run claude, followed by two hyphens

26
00:01:36,896 --> 00:01:39,938
and version, to check that it is available.

27
00:01:39,938 --> 00:01:43,955
You will need access through an eligible
Claude account or another supported

28
00:01:43,955 --> 00:01:48,495
authentication option. Follow the sign-in
prompts on first launch.

29
00:01:48,495 --> 00:01:52,964
For a subscription account, choose the Claude
sign-in option.

30
00:01:52,964 --> 00:01:57,771
API billing is separate. Now extract the
starter ZIP.

31
00:01:57,771 --> 00:02:03,692
In your terminal, type c d, a space, and the
path to that extracted folder.

32
00:02:03,692 --> 00:02:06,687
Put the path in quotes if it contains spaces.

33
00:02:06,687 --> 00:02:09,439
Run claude from inside that folder.

34
00:02:09,439 --> 00:02:14,210
Read the project trust prompt and continue
only for a folder you recognize.

35
00:02:14,210 --> 00:02:19,028
If the command is not found, reopen the
terminal and use the installation
troubleshooting

36
00:02:19,028 --> 00:02:24,114
link. There is no need to change your entire
computer setup just to complete

37
00:02:24,114 --> 00:02:25,112
this lesson.

38
00:02:25,167 --> 00:02:29,416
Before asking Claude to edit, ask it to
explain the project.

39
00:02:29,416 --> 00:02:34,049
Our starter contains a web page, a
stylesheet, a JavaScript file, and a short

40
00:02:34,049 --> 00:02:39,238
readme. Ask which file controls the
checkboxes and why selections disappear

41
00:02:39,238 --> 00:02:44,555
after a refresh. Keep the exploration focused
on this task.

42
00:02:44,555 --> 00:02:49,513
For an unfamiliar project, slash init can
help draft a CLAUDE dot M D instruction

43
00:02:49,513 --> 00:02:53,820
file. Review the draft and correct anything
it inferred incorrectly.

44
00:02:53,820 --> 00:02:57,222
The file is guidance, not an enforced
security boundary.

45
00:02:57,222 --> 00:03:02,156
For this exercise, useful instructions are
simple: preserve the visual design,

46
00:03:02,156 --> 00:03:06,719
use the existing plain JavaScript, and check
the reading list in a browser.

47
00:03:06,719 --> 00:03:11,305
A short, accurate file is more useful than a
long set of generic rules.

48
00:03:11,305 --> 00:03:14,602
Claude also has automatic memory for learned
patterns.

49
00:03:14,602 --> 00:03:18,166
That serves a different purpose from the
project rules you maintain.

50
00:03:18,166 --> 00:03:21,672
You do not need to configure an elaborate
memory system today.

51
00:03:21,672 --> 00:03:26,397
Start by understanding the files and making
your expectations explicit.

52
00:03:26,397 --> 00:03:31,018
If the project already has instructions, read
those first and avoid replacing

53
00:03:31,018 --> 00:03:32,783
them with a generic template.

54
00:03:32,833 --> 00:03:36,258
Now we turn the request into a small plan.

55
00:03:36,258 --> 00:03:40,414
In the terminal, Shift plus Tab cycles
through permission modes.

56
00:03:40,414 --> 00:03:43,247
Stop when the interface shows Plan mode.

57
00:03:43,247 --> 00:03:47,926
Confirm the displayed mode instead of
assuming one key press always lands in

58
00:03:47,926 --> 00:03:53,000
the right place. Ask Claude to propose how it
will save checked book identifiers

59
00:03:53,000 --> 00:03:55,450
and restore them on page load.

60
00:03:55,450 --> 00:04:00,477
A sensible plan for our project changes the
JavaScript, keeps the layout, and

61
00:04:00,477 --> 00:04:03,507
checks both selecting and clearing a book.

62
00:04:03,507 --> 00:04:08,766
Read the plan. If it adds a database, a new
framework, or a sign-in system,

63
00:04:08,766 --> 00:04:10,798
bring it back to the small task.

64
00:04:10,798 --> 00:04:14,025
Browser storage is enough for this practice
example.

65
00:04:14,025 --> 00:04:18,495
Once the plan matches your goal, leave Plan
mode and ask Claude to implement

66
00:04:18,495 --> 00:04:21,955
it. Read permission requests as they appear.

67
00:04:21,955 --> 00:04:26,680
A request to edit the expected project file
is different from a request to publish

68
00:04:26,680 --> 00:04:29,060
a site or delete a folder.

69
00:04:29,060 --> 00:04:33,402
If the scope is unclear, stop and ask for an
explanation.

70
00:04:33,402 --> 00:04:39,857
The useful habit is simple: explore, plan,
implement, verify.

71
00:04:39,857 --> 00:04:44,281
Planning is especially helpful when you are
new or the change has several moving

72
00:04:44,281 --> 00:04:45,047
parts.

73
00:04:45,100 --> 00:04:47,457
Claude says the change is complete.

74
00:04:47,457 --> 00:04:52,020
Now check it. Ask which files changed, what
the checks showed, and what it could

75
00:04:52,020 --> 00:04:57,198
not verify. Review the code changes in your
editor, or use git diff if your

76
00:04:57,198 --> 00:04:59,938
project is already tracked in Git.

77
00:04:59,938 --> 00:05:04,837
Our practice project does not require Git or
a build system.

78
00:05:04,837 --> 00:05:08,134
Open its index file in a browser.

79
00:05:08,134 --> 00:05:11,942
Check the first book, refresh, and confirm it
remains checked.

80
00:05:11,942 --> 00:05:16,888
Uncheck it, refresh again, and confirm it
stays cleared.

81
00:05:16,888 --> 00:05:20,348
Check a different book to make sure the items
are independent.

82
00:05:20,348 --> 00:05:22,542
Watch the browser console for errors.

83
00:05:22,542 --> 00:05:26,803
If your browser restricts storage for local
files, ask Claude to serve this

84
00:05:26,803 --> 00:05:30,193
folder locally and use that address
consistently.

85
00:05:30,193 --> 00:05:35,604
On an existing application, use its actual
build and test commands.

86
00:05:35,604 --> 00:05:40,921
Do not invent an npm test command when the
project has no test script.

87
00:05:40,921 --> 00:05:46,122
If a check fails, give Claude the exact
observed behavior and any relevant error

88
00:05:46,122 --> 00:05:50,395
text. Keep secrets out of pasted logs.

89
00:05:50,395 --> 00:05:54,156
Ask for a focused fix, then repeat the
affected checks.

90
00:05:54,156 --> 00:05:57,512
Tests, browser checks, and your review work
together.

91
00:05:57,512 --> 00:06:02,144
They do not guarantee that every possible bug
is gone, but they give you evidence

92
00:06:02,144 --> 00:06:03,909
about the change you asked for.

93
00:06:03,967 --> 00:06:09,052
As you work, the conversation fills with
instructions, file contents, and tool

94
00:06:09,052 --> 00:06:12,164
results. That is context.

95
00:06:12,164 --> 00:06:17,086
Keep requests focused and avoid dumping
unrelated files into the session.

96
00:06:17,086 --> 00:06:21,185
Slash context helps you inspect usage.

97
00:06:21,185 --> 00:06:25,991
Slash compact summarizes a long conversation,
and you can tell it which decisions

98
00:06:25,991 --> 00:06:31,274
and checks to preserve. When you move to an
unrelated task, slash clear starts

99
00:06:31,274 --> 00:06:35,325
a fresh conversation. Save any important
decisions first.

100
00:06:35,325 --> 00:06:39,366
Project instructions are still available when
the new conversation begins.

101
00:06:39,366 --> 00:06:44,057
There is no universal forty percent target
that guarantees better results.

102
00:06:44,057 --> 00:06:47,714
The practical question is whether the session
still has the information needed

103
00:06:47,714 --> 00:06:53,263
for the current task. If Claude starts
repeating an unsuccessful approach, stop,

104
00:06:53,263 --> 00:06:56,850
explain the mismatch, and narrow the next
step.

105
00:06:56,850 --> 00:07:01,634
Slash rewind can restore supported
conversation or file checkpoints, but it

106
00:07:01,634 --> 00:07:04,629
is not a general undo button.

107
00:07:04,629 --> 00:07:09,621
It cannot undo every shell command or an
external action, such as a published

108
00:07:09,621 --> 00:07:14,846
deployment. Keep normal backups and use Git
when it fits your project.

109
00:07:14,846 --> 00:07:20,291
Before ending today, ask Claude for a short
handoff: what changed,

110
00:07:20,291 --> 00:07:23,809
what passed, and what remains.

111
00:07:23,809 --> 00:07:27,048
That gives your next session somewhere
sensible to begin.

112
00:07:27,367 --> 00:07:29,701
You now have the core workflow.

113
00:07:29,701 --> 00:07:35,471
Define the result, inspect the project,
review a plan, implement a small change,

114
00:07:35,471 --> 00:07:40,660
and verify it. When you repeat a useful
procedure, a skill can package the
instructions

115
00:07:40,660 --> 00:07:45,583
in a SKILL dot M D file. Full skill
instructions load when used, though the

116
00:07:45,583 --> 00:07:49,403
skill listing itself can still consume
context.

117
00:07:49,403 --> 00:07:54,116
Subagents can investigate a focused question
in a separate context and return

118
00:07:54,116 --> 00:07:58,772
a summary. They are useful for independent
work, but every task does not need

119
00:07:58,772 --> 00:08:04,183
a team. M C P connects Claude to external
tools and data.

120
00:08:04,183 --> 00:08:08,374
Those connections bring their own permissions
and responsibilities, so add them

121
00:08:08,374 --> 00:08:13,146
for a specific need. None of these extensions
is required for your first successful

122
00:08:13,146 --> 00:08:18,626
session. Your next exercise is deliberately
small: ask Claude to show how

123
00:08:18,626 --> 00:08:23,699
many books remain unread, preserve the
design, and verify that the count changes

124
00:08:23,699 --> 00:08:26,033
when you select and clear books.

125
00:08:26,033 --> 00:08:29,260
Use the prompt builder to create a brief for
your own project.

126
00:08:29,260 --> 00:08:33,544
Then take the same habits with you: a clear
outcome, a sensible boundary, and

127
00:08:33,544 --> 00:08:35,274
evidence that the result works.

128
00:08:35,274 --> 00:08:37,445
That is a first session worth repeating.