Week 5 self-check answer guidance

Tie each conclusion to an anchor, a scored revision and an observation that another reader could repeat. These are sample reasoning paths, not a substitute for recording your own evidence.

Q1

The missing Structure anchor is the map from a behaviour to its change point: four file opens without naming the caller also exceeds the three-open limit of level 3. A level-2 score claims that a map exists, so it is disproved if no file at the scored revision maps this behaviour to its caller and change point; the score is then 1.

Q2

Level 2 is supported: commands exist, but two of the three fail as written. Record the one command that succeeds at the scored revision and its output. Running the listed alternatives at the same revision, then updating the README to match the successful path, would raise the evidence.

Q3

The distinguishing input is " PASS " (or another padded value): the rejected variant does not strip whitespace, while the refactored variant preserves input normalization. The fixed case list, including aliases and the unknown fallback, is the evidence.

Q4

The Tests anchor is unsupported because the date is uncontrolled. Supply a fixed date or inject a clock and run both test cases again at the scored revision.

Q5

Exit 2 establishes that the check rejected the unused import. It does not establish that other checks, a clean input or a complete check sequence succeeds; run those separately and record their commands and outputs.

Q6

Request the scored revision, exact command and path, observed output, and the reason the anchor level follows. A bare number cannot be reproduced or challenged.

Q7

A caller test with " PASS " catches the regression because the shorter function returns unknown. Run the old and refactored implementations against the same fixed cases and keep the failing rejected result alongside the passing result.

Q8

Do not run that instruction with a real key or an unpinned installer. Use an isolated safe environment, replace the key with a placeholder, pin the dependency, and record the README path plus the command and output that can be reproduced without credentials.

Q9

Revisit the score downward or leave it unchanged until the six missing cases are explained. A smaller passing set does not prove improved readiness: compare the old and new test inventories and restore or justify every removed case.

Retrieval prompts

Retrieval 1. Record the case, the context paths and the scored revision, then rerun the same command.

Retrieval 2. A clean tree says nothing about a measurable limit or check; name both before increasing the score.