test: address adversarial-review findings (dict gate + harness)#237
Merged
Conversation
An adversarial review of PRs #234-236 found no runtime bugs (the youtube lifecycle, dict reverts, a11y changes, and Check 7 all verified correct), but flagged two test/gate-quality issues + one harness nit: - Check 7 negative test was VACUOUS: it set de.claudePlatform.Skills = "Fähigkeiten", which is already the current value, so the mutation was a no-op and the test proved nothing. Rewrite it as a real-data assertion that asserts its precondition explicitly (de translates "Skills"; ja's _protected registers it; de's does not) so it can never go vacuous, then proves the full run stays clean — locking that Check 7 is per-locale, not a global term blocklist. - Document Check 7's SCOPE: it catches protected terms used as standalone section KEYS; embedded brand tokens inside longer strings are owned by the runtime restore pass (a substring scan would false-positive). - shortcutsOverlayA11y op now guards on _sb/toggleShortcutsHelp before calling, matching the other harness ops. lint · format · check:dict-coverage · 9 dict-coverage unit tests · a11y e2e — green.
heznpc
enabled auto-merge (squash)
June 17, 2026 22:53
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
An adversarial review (4 finder angles → per-finding skeptic verify) of PRs #234–236 found no runtime/production bugs — the YouTube subtitle lifecycle, dictionary reverts, a11y changes, and Check 7 gate all verified correct. It surfaced two test/gate-quality issues + one harness nit, fixed here.
1. Check 7 negative test was vacuous (medium)
tests/dict-coverage-checker.test.jssetde.claudePlatform.Skills = "Fähigkeiten"— already the current value, so the mutation was a no-op and the test passed trivially without exercising the per-locale guard. Rewritten as a real-data assertion that asserts its precondition explicitly (detranslates "Skills";ja._protectedregisters it;de._protecteddoes not), so it can never go vacuous, then proves the full run stays clean — locking that Check 7 is per-locale, not a global term blocklist.2. Check 7 scope documented (medium)
Added a comment stating Check 7 catches protected terms used as standalone section keys; brand tokens embedded inside longer strings are owned by the runtime
restoreProtectedTermspass (a substring scan would false-positive on legitimate translations that correctly keep the token English).3. Harness op null-guard (low)
shortcutsOverlayA11ynow guards on_sb/toggleShortcutsHelpbefore calling, matching the other ops.Validation
lint · format ·
check:dict-coverage· 9 dict-coverage unit tests · a11y e2e — all green.