Design review
gstack’s skill checks stop failing Opus 5.5 for valid work
“I rated this plan 4/10 for design completeness”
design focus checkno acknowledged Design focuscounts
Design reviewDocs syncQAYour setup
garrytan/gstack · CHANGELOG.md · 1.91.167.0
gstack's paid skill tests failed Opus 5.5 three times for a plan design review it ran correctly. A design completeness score out of ten now counts, however Opus words it.
Docs sync
The docs sync test accepts the reads Opus joins together
✓ cat personal-note.txt; git diff main HEAD
✓ for f in …; do cat $f; done
✓ git ls-files -z | tr '\0' '\n'
Still rejected: Git writes, and reading the Claude config dir, which holds credentials and every session transcript
Design reviewDocs syncQAYour setup
garrytan/gstack · CHANGELOG.md · 1.91.169.0
In the ship skill's docs sync test, joined Git reads, simple for loops and a pipe into tr now count as reads. Git writes still fail, and so does reading the Claude config folder.
QA
Three QA tests stop failing on valid agent work
qa-only accepts reads joined with its exact bookkeeping calls
ship stops counting one explicit plan check as a wasted rerun
Handoff check is one rule: read HANDOFF.md before you materialize or record a review
Design reviewDocs syncQAYour setup
garrytan/gstack · CHANGELOG.md · 1.91.168.0
The QA tests get the same treatment. qa-only accepts reads joined with its bookkeeping calls, one explicit plan check is no longer a wasted rerun, and the handoff check is one clear rule.
Your setup
gstack runs the same for you; its Opus 5.5 checks go red less often
CHANGELOG.md1.91.167.0 · 1.91.168.0 · 1.91.169.0Which model runs your gstack skills?
Design reviewDocs syncQAYour setup
garrytan/gstack · CHANGELOG.md · 1.91.167.0 to 1.91.169.0
All three releases change only gstack's tests, so it runs the same for you. The details are in the changelog. Which model runs your gstack skills?
