WEBVTT

1
00:00:00.000 --> 00:00:03.761
This challenge asks whether your
repository can carry a useful task

2
00:00:03.808 --> 00:00:07.581
into a new session. We will
add one meaningful assertion,

3
00:00:07.860 --> 00:00:10.762
verify it, and leave a
handoff with real receipts.

4
00:00:11.203 --> 00:00:13.514
Start with a fresh checkpoint-eight copy.

5
00:00:13.955 --> 00:00:17.809
If you already tried the negative-infinity
extension in an earlier lesson,

6
00:00:18.053 --> 00:00:20.584
preserve that work separately
and use the supplied

7
00:00:20.642 --> 00:00:23.452
starting checkpoint for
this reproducible exercise.

8
00:00:23.963 --> 00:00:26.529
Read the handoff before
opening the agent task.

9
00:00:26.854 --> 00:00:30.116
It asks for an assertion proving
that negative infinity produces

10
00:00:30.197 --> 00:00:33.204
fault with a nonfinite error
and no current temperature.

11
00:00:33.529 --> 00:00:36.107
Production behavior
should remain unchanged.

12
00:00:36.385 --> 00:00:40.727
The normal and fault requirements already
promise that nonfinite input is invalid,

13
00:00:40.936 --> 00:00:44.059
so this task strengthens
verification of an existing

14
00:00:44.129 --> 00:00:46.741
contract rather than adding a new feature.

15
00:00:47.345 --> 00:00:50.201
Check the starting project
status and source identity.

16
00:00:50.584 --> 00:00:55.344
The expected stage has normal and fault
behavior, replay, and basic assertions.

17
00:00:55.855 --> 00:00:59.988
It has no configurable threshold,
hysteresis, or command parser.

18
00:01:00.313 --> 00:01:02.669
If the task starts
discussing those features,

19
00:01:02.844 --> 00:01:06.895
it may be reading the wrong checkpoint or
broadening the work beyond the handoff.

20
00:01:07.221 --> 00:01:10.436
Establishing the current state
prevents that confusion early.

21
00:01:10.819 --> 00:01:14.012
Start a genuinely fresh agent
session in the project root.

22
00:01:14.256 --> 00:01:17.449
Ask it to complete the handoff
using repository context.

23
00:01:17.692 --> 00:01:21.326
Do not paste the solution or exact
command sequence into the prompt.

24
00:01:21.570 --> 00:01:25.785
A continued conversation can already
contain those details, which would weaken

25
00:01:25.831 --> 00:01:29.964
the test of whether the repository
supplies the information needed to resume.

26
00:01:30.406 --> 00:01:33.041
Observe actual context inspection.

27
00:01:33.482 --> 00:01:36.884
The task should find the handoff,
project guidance, README,

28
00:01:36.965 --> 00:01:39.786
applicable requirements,
and relevant test source.

29
00:01:40.170 --> 00:01:44.350
A final summary that sounds informed
does not prove which files were used.

30
00:01:44.791 --> 00:01:47.473
Preserve the available
file-read or tool evidence.

31
00:01:47.914 --> 00:01:50.758
If the interface does not
expose part of that process,

32
00:01:51.002 --> 00:01:55.449
record the limit rather than inventing
it. Now inspect the proposed test.

33
00:01:55.959 --> 00:01:58.572
Negative infinity is a nonfinite value,

34
00:01:58.723 --> 00:02:01.788
and the expected state is
fault with a nonfinite error.

35
00:02:01.997 --> 00:02:05.735
The current flag must be false.
A separate monitor instance

36
00:02:05.770 --> 00:02:09.195
is a simple way to isolate the
case from earlier recovery state.

37
00:02:09.520 --> 00:02:12.667
Carefully ordered injected
timestamps can also work,

38
00:02:12.875 --> 00:02:17.125
but the assertion should not accidentally
depend on the preceding test's history.

39
00:02:17.636 --> 00:02:20.956
The distinction between policy
and output matters here.

40
00:02:21.560 --> 00:02:24.718
The new assertion checks state,
error, and availability.

41
00:02:25.228 --> 00:02:28.352
Existing output checks
handle how an unavailable

42
00:02:28.386 --> 00:02:31.150
temperature is serialized as JSON null.

43
00:02:31.475 --> 00:02:35.678
Do not claim the new policy assertion
alone tested every output field.

44
00:02:36.003 --> 00:02:38.952
Each check should have a clear
purpose so later maintenance

45
00:02:38.975 --> 00:02:42.272
can preserve its meaning.
Watch the scope of the edit.

46
00:02:42.447 --> 00:02:45.407
The intended result is a small
addition to the test harness,

47
00:02:45.616 --> 00:02:49.714
plus an appropriate evidence or handoff
update. A large production refactor

48
00:02:49.831 --> 00:02:53.488
is not required to add coverage
for behavior that already exists.

49
00:02:53.813 --> 00:02:56.007
If the new test unexpectedly fails,

50
00:02:56.251 --> 00:02:59.362
preserve the discrepancy and
diagnose it against the requirement

51
00:02:59.409 --> 00:03:03.403
before deciding whether production code
needs a separate correction. The task

52
00:03:03.449 --> 00:03:05.806
should run the relevant
documented test command,

53
00:03:05.980 --> 00:03:09.045
and any additional check
justified by the actual change.

54
00:03:09.556 --> 00:03:12.412
An announced intention
to run tests is a plan.

55
00:03:12.795 --> 00:03:16.162
A command with an actual
process result is execution.

56
00:03:16.545 --> 00:03:20.167
Save the output and source
identity. A confident sentence

57
00:03:20.190 --> 00:03:24.172
that tests passed is not enough when
there is no accessible receipt behind it.

58
00:03:24.555 --> 00:03:27.969
Independently review the diff
after the agent finishes.

59
00:03:28.294 --> 00:03:32.148
Confirm that the expected values
come from the nonfinite requirement

60
00:03:32.322 --> 00:03:37.036
and that unrelated files remain
untouched. Read the actual test result.

61
00:03:37.361 --> 00:03:41.691
If you perform an additional independent
run, record it as your verification.

62
00:03:41.935 --> 00:03:45.952
It can establish the artifact's current
behavior, but it does not retroactively

63
00:03:46.010 --> 00:03:49.087
prove the agent executed a
command it only announced.

64
00:03:49.470 --> 00:03:53.649
Suppose the handoff claims that earlier
checks passed but points to no evidence.

65
00:03:53.975 --> 00:03:55.879
Search the appropriate result location.

66
00:03:56.122 --> 00:03:59.269
If the receipt cannot be found,
mark the claim unestablished.

67
00:03:59.594 --> 00:04:02.717
A new run can provide current
evidence, but the record

68
00:04:02.763 --> 00:04:06.281
should not pretend to recreate an
earlier unrecorded observation.

69
00:04:06.722 --> 00:04:10.530
This is a common place where an
unsupported claim can otherwise be copied

70
00:04:10.588 --> 00:04:14.431
indefinitely. Update the handoff
with the current source identity,

71
00:04:14.640 --> 00:04:18.100
the completed bounded test change,
relevant command results,

72
00:04:18.274 --> 00:04:20.979
remaining limitations,
and one next action.

73
00:04:21.362 --> 00:04:25.170
The capstone can be named as
future work. Do not describe

74
00:04:25.194 --> 00:04:28.816
its alert or parser behavior as already
implemented in checkpoint eight.

75
00:04:29.199 --> 00:04:33.901
A concise accurate handoff is more useful
than a long summary that mixes stages.

76
00:04:34.342 --> 00:04:38.325
Your challenge submission includes the
fresh prompt, actual task evidence,

77
00:04:38.476 --> 00:04:42.284
changed assertion, reviewed diff,
command results, independent review,

78
00:04:42.342 --> 00:04:46.010
and updated handoff. Use the answer
key after attempting the work.

79
00:04:46.254 --> 00:04:49.087
Grade the outcome and evidence
rather than exact wording.

80
00:04:49.331 --> 00:04:52.350
Another model may explain the
same correct action differently,

81
00:04:52.466 --> 00:04:55.682
and a polished explanation can
still lack a required action.

82
00:04:56.193 --> 00:04:58.201
Keep the assertion's input visible.

83
00:04:58.526 --> 00:05:01.974
Use the standard numeric-limits
expression for negative infinity

84
00:05:02.358 --> 00:05:06.467
rather than a magic bit pattern that
depends on an unstated representation.

85
00:05:06.711 --> 00:05:09.045
The requirement is about
a nonfinite numeric

86
00:05:09.103 --> 00:05:12.191
value, and the test should
communicate that intent directly.

87
00:05:12.574 --> 00:05:16.394
A reader should understand the case
without decoding a special constant.

88
00:05:16.719 --> 00:05:19.831
If an account limit or missing
tool prevents execution,

89
00:05:20.074 --> 00:05:22.594
preserve the prepared
task and static review.

90
00:05:23.035 --> 00:05:26.831
Mark the runtime gate open and
identify the exact dependency.

91
00:05:27.273 --> 00:05:30.883
Do not substitute the supplied reference
result for your own missing run.

92
00:05:31.208 --> 00:05:34.901
The written work can remain useful
while the behavior test is incomplete,

93
00:05:35.028 --> 00:05:37.373
and the record should
make those statuses clear.

94
00:05:37.699 --> 00:05:41.182
Pause here to complete the challenge
and then take the module quiz.

95
00:05:41.507 --> 00:05:45.512
The recommended quiz target is
four correct answers out of five,

96
00:05:45.756 --> 00:05:48.275
with unlimited retry and explanations.

97
00:05:48.554 --> 00:05:51.561
Quiz progress is separate
from the practical evidence.

98
00:05:52.257 --> 00:05:56.390
A completion button cannot establish that
the test was meaningful or that the fresh

99
00:05:56.437 --> 00:05:58.677
session used the repository correctly.

100
00:05:59.374 --> 00:06:02.253
When you finish, explain
what the handoff achieved.

101
00:06:02.764 --> 00:06:06.584
A successful run found the right
task from repository context,

102
00:06:06.735 --> 00:06:09.069
made a bounded
requirement-based test change,

103
00:06:09.185 --> 00:06:13.179
performed relevant verification, and left
current evidence for the next session.

104
00:06:13.504 --> 00:06:15.663
That is a practical form of continuity.

105
00:06:15.988 --> 00:06:19.610
You are now ready to apply the same
discipline to the larger capstone

106
00:06:19.726 --> 00:06:23.267
without losing the boundaries that
made this small task reviewable.