WEBVTT

1
00:00:00.000 --> 00:00:03.704
This challenge brings together
project guidance, source discipline,

2
00:00:03.855 --> 00:00:07.907
skill packaging, and behavior
testing. You will deliver a reusable

3
00:00:07.941 --> 00:00:11.401
constraints procedure and explain
why its boundaries are useful.

4
00:00:11.842 --> 00:00:16.370
Start with a new checkpoint-four exercise
copy and preserve your previous version.

5
00:00:16.753 --> 00:00:20.456
The aim is a small supported
improvement, not a complete redesign

6
00:00:20.491 --> 00:00:24.590
of everything you wrote in the module.
Review the skill's purpose before editing.

7
00:00:24.915 --> 00:00:28.537
It derives constraints for an identified
board or component from primary

8
00:00:28.595 --> 00:00:33.007
technical material. It needs exact
identity, relevant source sections,

9
00:00:33.181 --> 00:00:36.246
units and conditions, and
explicit missing information.

10
00:00:36.629 --> 00:00:40.518
It should not turn an extraction
request into unrelated firmware edits,

11
00:00:40.727 --> 00:00:42.852
purchases, or physical operations.

12
00:00:43.177 --> 00:00:46.369
Those boundaries keep the output
aligned with the actual task.

13
00:00:46.613 --> 00:00:48.564
Check the division of responsibility.

14
00:00:49.075 --> 00:00:52.790
AGENTS dot md holds recurring
project expectations.

15
00:00:53.115 --> 00:00:55.669
The hardware reference
holds current sourced facts.

16
00:00:55.878 --> 00:00:58.153
The skill holds the extraction procedure.

17
00:00:58.362 --> 00:01:01.149
The dependency record
preserves target and tool pins.

18
00:01:01.474 --> 00:01:04.922
If the same detailed fact has
been copied into all four places,

19
00:01:05.200 --> 00:01:08.416
consider whether a link would make
the context easier to maintain.

20
00:01:08.695 --> 00:01:12.329
Avoid creating a new contradiction
while trying to improve reuse.

21
00:01:12.654 --> 00:01:15.707
Choose one weakness demonstrated
by your previous tests.

22
00:01:16.032 --> 00:01:19.678
Perhaps the relevant request did not
select the skill. Perhaps a spelling

23
00:01:19.713 --> 00:01:23.520
correction triggered a constraints table.
Perhaps incomplete source material

24
00:01:23.590 --> 00:01:27.781
produced an invented pin. Perhaps the
supporting reference path was broken.

25
00:01:28.106 --> 00:01:32.147
Each case suggests a different repair.
Start with the observed failure

26
00:01:32.298 --> 00:01:36.675
rather than adding a long list of generic
instructions. For a broad trigger,

27
00:01:36.849 --> 00:01:39.786
refine the description around
the actual input and task.

28
00:01:40.065 --> 00:01:43.281
Deriving interface or operating
constraints from supplied primary

29
00:01:43.339 --> 00:01:47.275
documents is specific. Helping
with anything embedded is not.

30
00:01:47.553 --> 00:01:51.222
After the edit, the positive case must
still be able to select the skill,

31
00:01:51.304 --> 00:01:54.682
and the unrelated case should
remain ordinary prose work.

32
00:01:54.961 --> 00:01:59.152
A description that prevents all
activation is not a useful solution.

33
00:01:59.477 --> 00:02:03.018
For an unsupported row, improve
the procedure's source review.

34
00:02:03.343 --> 00:02:07.302
Require each factual value to be
checked against its cited section.

35
00:02:07.627 --> 00:02:12.248
Add an example where the exact pin mapping
is absent and must remain unresolved.

36
00:02:12.631 --> 00:02:16.450
Explain that the next step is to
obtain the applicable board reference,

37
00:02:16.625 --> 00:02:19.365
not to supply a familiar
value from another variant.

38
00:02:19.748 --> 00:02:24.787
Keep uncertainty actionable rather than
merely apologetic. For a broken reference,

39
00:02:24.903 --> 00:02:28.595
correct the relative path and open
the file from the skill's location.

40
00:02:28.978 --> 00:02:32.507
The focused reference should add
something that helps the procedure, such

41
00:02:32.542 --> 00:02:36.396
as source identity, units,
conditions, and missing-data examples.

42
00:02:36.907 --> 00:02:40.692
A large unrelated document does
not make the skill more capable.

43
00:02:41.203 --> 00:02:44.163
It can make the relevant steps
harder to find and maintain.

44
00:02:44.547 --> 00:02:47.496
Once the change is clear,
run the relevant case again.

45
00:02:47.937 --> 00:02:52.174
Use the product five four seven
seven and BME280 source path

46
00:02:52.221 --> 00:02:56.516
in the hardware reference. Explicit
invocation can diagnose discovery,

47
00:02:56.691 --> 00:02:59.848
but the implicit case must
also be evaluated separately.

48
00:03:00.174 --> 00:03:02.496
Preserve the exact prompt, source scope,

49
00:03:02.704 --> 00:03:06.001
skill version, host version,
and available selection evidence.

50
00:03:06.327 --> 00:03:09.368
Do not replace the actual output
with the table you expected.

51
00:03:09.751 --> 00:03:12.248
Review the constraints table row by row.

52
00:03:12.491 --> 00:03:16.810
Identify the object each claim describes:
chip, breakout, or host board.

53
00:03:17.193 --> 00:03:19.457
Check the value, unit, and source section.

54
00:03:19.899 --> 00:03:22.441
Look for conditions that
were lost during extraction.

55
00:03:22.882 --> 00:03:26.586
Then inspect the consequence. The
consequence should explain what

56
00:03:26.621 --> 00:03:28.223
the fact means for this project,

57
00:03:28.258 --> 00:03:32.019
and any inference should be labeled so
a reviewer can examine the reasoning.

58
00:03:32.716 --> 00:03:36.477
Run the unrelated case in a fresh
task without mentioning the skill.

59
00:03:37.290 --> 00:03:40.808
The spelling correction should stay
a spelling correction. Then run

60
00:03:40.854 --> 00:03:44.964
the incomplete-input case for a different
board from the documented example,

61
00:03:45.115 --> 00:03:47.867
with no exact variant
or schematic supplied.

62
00:03:48.308 --> 00:03:51.988
Use the exact course prompt and its
description-only source boundary.

63
00:03:52.429 --> 00:03:55.448
Missing identity and source
evidence should remain explicit.

64
00:03:55.831 --> 00:03:58.907
These cases test useful
restraint as well as capability.

65
00:03:59.233 --> 00:04:02.437
A skill that produces an impressive
answer by guessing has not met

66
00:04:02.472 --> 00:04:06.233
the evidence contract. Compare the
new records with the previous ones.

67
00:04:06.617 --> 00:04:10.041
Explain what changed and which
observation supports the improvement.

68
00:04:10.552 --> 00:04:14.627
Keep the failed attempt when it motivated
the repair. If an account or tool limit

69
00:04:14.685 --> 00:04:18.749
prevents the rerun, record the open
gate and preserve the complete package.

70
00:04:19.191 --> 00:04:21.559
A planned evaluation
is useful preparation,

71
00:04:21.838 --> 00:04:24.369
but it must not be promoted
to an observed pass.

72
00:04:24.694 --> 00:04:28.188
Prepare a concise handoff.
Name the skill's location,

73
00:04:28.339 --> 00:04:31.462
how it was invoked in the
tested host, which cases passed,

74
00:04:31.543 --> 00:04:35.061
which remain open, and what should
be rerun after a relevant change.

75
00:04:35.874 --> 00:04:39.032
Link to the actual case records.
The next session should be

76
00:04:39.067 --> 00:04:41.644
able to inspect the current
procedure and continue

77
00:04:41.667 --> 00:04:45.174
from the unresolved question without
recovering your old conversation.

78
00:04:46.126 --> 00:04:50.294
Name one remaining physical unknown.
A sourced constraints table

79
00:04:50.329 --> 00:04:54.613
can establish what a manufacturer
document says about the selected example.

80
00:04:54.996 --> 00:04:58.966
It does not establish the revision of a
board you have not inspected, the position

81
00:04:59.001 --> 00:05:02.554
of a jumper, successful wiring,
or actual measurement accuracy.

82
00:05:02.833 --> 00:05:05.689
Physical hardware remains
optional for the core course,

83
00:05:05.898 --> 00:05:09.184
and its evidence belongs in a
separate clearly labeled activity.

84
00:05:09.567 --> 00:05:12.934
Your submission contains the
final skill, focused reference,

85
00:05:13.108 --> 00:05:17.624
any bounded instruction correction, the
four actual case records, and the handoff.

86
00:05:18.228 --> 00:05:21.189
Use the rubric in the answer
key after your own attempt.

87
00:05:21.630 --> 00:05:25.995
Strong evidence matters more than
reproducing exact wording. A package

88
00:05:26.018 --> 00:05:29.826
with valid metadata but no behavior
record is incomplete for this challenge,

89
00:05:30.000 --> 00:05:33.820
while an honest blocked record tells
us precisely what remains to be done.

90
00:05:34.261 --> 00:05:36.119
Take a moment to consider transfer.

91
00:05:36.560 --> 00:05:39.567
The same structure can support
a bounded firmware review,

92
00:05:39.741 --> 00:05:42.910
a test-evidence summary,
or a log-triage procedure.

93
00:05:43.351 --> 00:05:47.148
Start with the inputs and output, then
define what would make a result wrong.

94
00:05:47.531 --> 00:05:50.596
Choose relevant, unrelated,
and incomplete cases.

95
00:05:50.979 --> 00:05:52.929
If you cannot state failure conditions,

96
00:05:53.080 --> 00:05:56.157
the procedure is probably too
broad to package confidently.

97
00:05:56.598 --> 00:06:00.337
Finish with the module quiz and read
the explanations for missed questions.

98
00:06:00.778 --> 00:06:04.737
The recommended target is four
out of five with unlimited retry.

99
00:06:05.178 --> 00:06:08.522
Quiz progress remains separate
from the practical evidence.

100
00:06:08.905 --> 00:06:12.701
You now have a repeatable way to turn an
engineering procedure into a maintained

101
00:06:12.771 --> 00:06:17.009
skill while keeping sources, uncertainty,
and observed behavior visible.

102
00:06:17.450 --> 00:06:21.037
Chapter Sixteen will use that pattern
for the second required skill,

103
00:06:21.211 --> 00:06:22.337
firmware review.