Opus 5.5
gstack now tests every skill change on Claude Opus 5.5
Fable 5.1Opus 5.5gstack's paid pull request gate now tests every skill change on Claude Opus 5.5 instead of Fable 5.1. The first full Opus gate passed all a hundred and thirty-two of its cases.
Cost
A full gate costs about a third of Fable’s
A full gate now costs about a third as much, thirty-two dollars against Fable's ninety-three. Sessions take about half the time, with Opus at medium effort by default and Fable at high.
Grading
The judge stays Fable 5.1
Grading stays put. The judge is still Fable 5.1, because its calibration is keyed to it. Opus results start a new pass rate series, so they never mix with Fable's.
Opus wording
The gate now understands how Opus phrases things
The checks also learned how Opus talks. Saying integrate the hook yourself now counts as manual setup, and a continue or abort question is recognized by its options. An unanswered permission dialog fails within twenty seconds.
Your runs
Need Fable for one run? The changelog covers it
CHANGELOG.mdgstack release notes, “If you…”Need Fable for a single run, or comparing pass rates across the switch? The changelog covers both. Which model do your gstack skills run on, Opus 5.5 or Fable 5.1?
