[polewskm]

Pins two behaviors that only main() held, so tests now reach them, with no behavior or log text changed:

• Test that judge spend stays out of the eval review cost #107: the evals' per-case tally moves into tallyCase in evals/run.ts, with tests that review cost leaves out the concern and precision judges' calls while the case cost counts them.
• Test /review explain's answer() directly #108: writeForPublish moves verbatim to src/outputs.ts, and the explain answer's writing moves into writeAnswer in src/explain.ts, with tests that an unknown finding writes only answer.json without a model call, trace.json is written when the call ran, answered is set, and results.json is never written.
• The branch merges main in to resolve a CHANGELOG.md conflict, so merge this with a merge commit, not by rebasing.

Closes #107
Closes #108

A first review round tested that writeAnswer refuses a secret in the answer or trace; main is merged in after #104.

[ncode-review]

🤖 NCode Review at d52b787: 2 new findings on the changed lines.

[ncode-review]

_{Superseded by the review at 5d49e7e.}

[polewskm]

/review

[ncode-review]

🤖 Review complete

Verdict: ✅ Approve — Re-review of what changed since the last review at d52b787: 1 hunk in 1 file, by correctness only. Earlier findings: F1 fixed. No further findings above nits.

1 reviewer run — 0 findings (run details)

Reviewer	Model	Findings	Top
correctness	sonnet-5-5	0	🟢

_{Trial: findings are posted as review threads and in the check run. $0.03 on Amazon Bedrock, 8s, at 5d49e7e. (how to interact)}

📋 Outcomes

Merged at 5d49e7e: 1 fixed, 1 dismissed, 0 acknowledged, 0 ignored.

Finding	Reviewer	Severity	Outcome
F1	tests	🟡 Nit	👎 dismissed
F2	hygiene	🟡 Nit	✅ fixed