Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prepare a release PR by bumping package versions (patch, minor, or major).
.claude/skills/itsbencodes-prepare-release/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -2% | 0% |
Ask the user for the bump type (patch | minor | major) if not provided.
Each publishable package is bumped independently — only packages that have changed since the last release tag are bumped (and later published). Packages may drift in version over time; this is fine because cross-package dependencies use caret ranges.
bashgit fetch origin main --tags
bashLAST_TAG=$(git describe --tags --abbrev=0 origin/main) echo "Last release: $LAST_TAG"
For each publishable package (skip react-cookie-demo), check whether its directory has changed since $LAST_TAG:
bashBUMP_TYPE='<patch|minor|major>' CHANGED=() for pkg in react-cookie universal-cookie universal-cookie-express universal-cookie-koa; do if ! git diff --quiet "$LAST_TAG" origin/main -- "packages/$pkg/"; then CHANGED+=("$pkg") fi done if [ ${#CHANGED[@]} -eq 0 ]; then echo "No packages changed since $LAST_TAG — nothing to release." exit 1 fi echo "Changed packages: ${CHANGED[*]}"
Each changed package bumps from its own current version on origin/main. The release version (used for the branch name and PR title) is the highest new version across all bumped packages.
bashRELEASE_VERSION="" declare -A NEW_VERSIONS for pkg in "${CHANGED[@]}"; do NV=$(node -p " const v = JSON.parse(require('child_process').execSync('git show origin/main:packages/$pkg/package.json')).version.split('.').map(Number); const t = '$BUMP_TYPE'; if (t==='major') { v[0]++; v[1]=0; v[2]=0; } else if (t==='minor') { v[1]++; v[2]=0; } else { v[2]++; } v.join('.') ") NEW_VERSIONS[$pkg]=$NV if [ -z "$RELEASE_VERSION" ] || [ "$(printf '%s\n%s' "$NV" "$RELEASE_VERSION" | sort -V | tail -1)" = "$NV" ]; then RELEASE_VERSION=$NV fi done echo "Release version: v$RELEASE_VERSION" for pkg in "${CHANGED[@]}"; do echo " $pkg -> ${NEW_VERSIONS[$pkg]}"; done
bashgit checkout -b release/v$RELEASE_VERSION origin/main
bashfor pkg in "${CHANGED[@]}"; do NV=${NEW_VERSIONS[$pkg]} node -e "const f='packages/$pkg/package.json';const p=require('./'+f);p.version='$NV';require('fs').writeFileSync(f,JSON.stringify(p,null,2)+'\n')" done
Verify with grep -H '"version"' packages/*/package.json.
Use the commit skill. Stage only packages/*/package.json. Title Bump to v$RELEASE_VERSION. Body lists each bumped package:
Release v$RELEASE_VERSION
- <pkg> -> <new-version>After PR is open, enable auto-merge:
bashgh pr merge --auto --squash
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,531 | 7,762 | -60% | 1 | 1 | 0% | 240 | 1,075 | +348% | 0 | 0 | — |
case-02 | fail→fail | 7,187 | 4,776 | -34% | 1 | 1 | 0% | 343 | 1,051 | +206% | 0 | 0 | — |
case-03 | fail→fail | 6,526 | 8,410 | +29% | 1 | 1 | 0% | 369 | 1,185 | +221% | 0 | 0 | — |
case-04 | fail→fail | 13,581 | 8,581 | -37% | 1 | 1 | 0% | 2,222 | 1,237 | -44% | 0 | 0 | — |
case-05 | pass→pass | 12,163 | 6,449 | -47% | 1 | 1 | 0% | 1,790 | 1,811 | +1% | 0 | 0 | — |
case-06 | pass→pass | 16,642 | 14,715 | -12% | 1 | 1 | 0% | 2,812 | 3,367 | +20% | 0 | 0 | — |
case-07 | fail→pass | 13,215 | 3,811 | -71% | 1 | 1 | 0% | 2,094 | 1,544 | -26% | 0 | 0 | — |
case-08 | pass→pass | 14,805 | 6,088 | -59% | 1 | 1 | 0% | 2,148 | 1,326 | -38% | 0 | 0 | — |
case-09 | fail→pass | 7,456 | 3,280 | -56% | 1 | 1 | 0% | 935 | 1,309 | +40% | 0 | 0 | — |
case-10 | pass→pass | 10,959 | 3,382 | -69% | 1 | 1 | 0% | 1,546 | 1,357 | -12% | 0 | 0 | — |
case-11 | fail→pass | 18,517 | 1,948 | -89% | 1 | 1 | 0% | 3,163 | 1,120 | -65% | 0 | 0 | — |
case-12 | fail→pass | 13,614 | 3,125 | -77% | 1 | 1 | 0% | 1,967 | 1,365 | -31% | 0 | 0 | — |
case-13 | pass→pass | 17,325 | 2,359 | -86% | 1 | 1 | 0% | 2,539 | 1,105 | -56% | 0 | 0 | — |
case-14 | pass→pass | 5,270 | 1,712 | -68% | 1 | 1 | 0% | 805 | 1,045 | +30% | 0 | 0 | — |
case-15 | pass→pass | 2,880 | 2,806 | -3% | 1 | 1 | 0% | 504 | 1,215 | +141% | 0 | 0 | — |
case-16 | pass→pass | 2,943 | 2,727 | -7% | 1 | 1 | 0% | 491 | 1,284 | +162% | 0 | 0 | — |
case-17 | pass→pass | 3,883 | 3,653 | -6% | 1 | 1 | 0% | 451 | 1,278 | +183% | 0 | 0 | — |
case-18 | fail→fail | 18,581 | 2,629 | -86% | 1 | 1 | 0% | 2,415 | 1,125 | -53% | 0 | 0 | — |
case-19 | fail→pass | 7,422 | 3,013 | -59% | 1 | 1 | 0% | 1,238 | 1,210 | -2% | 0 | 0 | — |
case-20 | fail→pass | 9,483 | 2,548 | -73% | 1 | 1 | 0% | 1,351 | 1,252 | -7% | 0 | 0 | — |
case-21 | pass→pass | 6,457 | 2,539 | -61% | 1 | 1 | 0% | 1,151 | 1,165 | +1% | 0 | 0 | — |
case-22 | pass→pass | 11,425 | 5,481 | -52% | 1 | 1 | 0% | 1,645 | 1,556 | -5% | 0 | 0 | — |
case-23 | pass→pass | 13,251 | 7,870 | -41% | 1 | 1 | 0% | 2,100 | 2,144 | +2% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 19 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +26 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.