fix(regression): judge coverage + fault classification (#41 F1/F2/F3) - #94
Merged
Conversation
From the independent re-verification of #41's judge: - F1 (Important): the count check validated cardinality, not coverage — a judge returning the right NUMBER of entries but all numbered n:1 (or n:99) all-pass passed the scenario while criteria 2..N were never graded (the exact skip-the-failing-criterion threat H1 targets). parseVerdict now requires the criteria to be numbered exactly {1..expectedCount}, each once, else a criteria-coverage-mismatch fault. - F3: a structurally malformed entry (no valid verdict) was recomputed to a fail (exit 1, a skill regression) — now a malformed-criterion fault (exit 2), so a broken judge is never mislabeled as a regression. - F2: judge.sh swallowed a failed --count into EXPECTED="" and graded "of 0"; it now fails loudly (exit 2) on a count-lookup failure or non-positive count. Tests: F1 coverage (duplicate/out-of-range n rejected, complete 1..N passes), F3 fault, and the existing malformed test updated to expect the fault. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KwyFyRC6pepaK6DuMK7YeT
huiyu
added a commit
that referenced
this pull request
Aug 24, 2026
…resolved PRs #90-#94 merged to 3.0; the 3 CRITICAL (C1 #62 writer, C2/C3 #51 claude path) that the independent pass overturned are re-fixed with real end-to-end tests, and the 18 further Important/Medium/Minor defects are closed. Residual risks R1-R4 remain post-3.0 by intent. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KwyFyRC6pepaK6DuMK7YeT
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.
Campaign-2 fix #5 (final) — #41 judge. From the independent re-verification.
n:1(orn:99), all-pass, passed the scenario while criteria 2..N were never graded (the exact skip-the-failing-criterion threat H1 targets).parseVerdictnow requires the criteria to be numbered exactly{1..expectedCount}, each once, else acriteria-coverage-mismatchfault.fail(exit 1, a skill regression) → now amalformed-criterionfault (exit 2), so a broken judge is never mislabeled as a regression.judge.shswallowed a failed--countintoEXPECTED=""and graded "of 0" → now fails loudly (exit 2) on a count-lookup failure or non-positive count.Verification
test:regression-judge: 15/15 — F1 (duplicate/out-of-rangenrejected, complete 1..N passes), F3 fault, existing malformed test updated to expect the fault. Shell logic smoke-tested (valid→count, unknown→exit 2). No vendored files changed.🤖 Generated with Claude Code