$ skilled compare passing.json regressed.json --fail-on-regression
Compared 1 matched cases: 1 regressions, 0 improvements.
Coverage: 0 added, 0 removed. 1 status changes; 1 comparable scores.
Regression policy: PASS to any non-PASS status, or a decrease in a comparable assertion pass rate.
"01-day-zero-full" ["with_skill"] changed: PASS -> FAIL; score 1 -> 0 (delta -1) REGRESSION
Warning: Reports do not contain complete evaluation-input, skill, fixture and judge-configuration fingerprints. Matching case IDs and prompts do not prove equivalent inputs or grading conditions.
Error: comparison found 1 regression(s)

Exit code: 1