Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Harvest one community PR into a release branch with authorship and credit preserved, verified green, and a warm thank-you.
.claude/skills/hmbown-gh-credit-harvest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 223% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 69% | 0% |
Harvest exactly one community PR into the real landing branch with full authorship and machine-readable credit, verified green, then thank the contributor. A PR is evidence: judge it from code, tests, comments, and checks, never the title. Do not merge, close, tag, or publish without maintainer approval — this skill lands a credited commit and posts thanks; the workflow closes the PR.
gh-close-issues).<release-branch>); a main-based "mergeable" flag does not prove it lands cleanly.main) and fetch the PR head:bash git switch <release-branch> git fetch origin pull/<N>/head git log -1 --format='%H %an <%ae>' FETCH_HEAD # author to preserve
bash gh pr view <N> --repo Hmbown/CodeWhale --json title,author,files,statusCheckRollup gh pr diff <N> --repo Hmbown/CodeWhale
bash git merge-tree $(git merge-base HEAD FETCH_HEAD) HEAD FETCH_HEAD # empty/clean = no conflict
bash git cherry-pick <sha> # one or more commits from FETCH_HEAD
--author and the co-author from .github/AUTHOR_MAP (fall back to numeric noreply):bash gh api users/<handle> --jq '"\(.id)+\(.login)@users.noreply.github.com"' git commit --author="Name <ID+handle@users.noreply.github.com>" -m "fix(scope): what changed (#<N>)" \ -m "Harvested from PR #<N> by @<handle>" \ -m "Co-authored-by: Name <ID+handle@users.noreply.github.com>" The Harvested from PR #<N> by @<handle> line in the body is what .github/workflows/auto-close-harvested.yml matches to auto-close with credit once the commit reaches main.
bash cargo fmt --all cargo test -p <crate> # the crate(s) the PR touched, not the whole workspace git log -1 --format='%an <%ae>%n%(trailers:key=Co-authored-by)' # credit the human, from .github/AUTHOR_MAP
main:bash gh pr comment <N> --repo Hmbown/CodeWhale \ --body "Thank you @<handle> — clean fix for <the specific bug>. Harvested into the <release-branch> lane with your authorship preserved; it'll auto-close with credit once it reaches main."
Shape of a clean harvest: a single-commit PR that cherry-picks with no conflicts needs no manual trailers — the author is preserved automatically — and a focused cargo test -p on the touched crate is enough to land it green.
git merge-tree.--author + trailers when you must..github/AUTHOR_MAP, then numeric noreply; never raw third-party, .local, placeholder, or bot emails.Harvested from PR #<N> by @<handle> body line — without it the PR won't auto-close with credit.gh-close-issues.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,616 | 14,145 | +33% | 1 | 1 | 0% | 256 | 1,850 | +623% | 0 | 0 | — |
case-02 | fail→fail | 8,039 | 11,317 | +41% | 1 | 1 | 0% | 238 | 1,612 | +577% | 0 | 0 | — |
case-03 | fail→fail | 11,472 | 9,581 | -16% | 1 | 1 | 0% | 277 | 1,385 | +400% | 0 | 0 | — |
case-04 | fail→pass | 12,372 | 7,863 | -36% | 1 | 1 | 0% | 1,982 | 2,251 | +14% | 0 | 0 | — |
case-05 | pass→pass | 21,188 | 3,882 | -82% | 1 | 1 | 0% | 2,287 | 1,601 | -30% | 0 | 0 | — |
case-06 | fail→pass | 13,676 | 6,689 | -51% | 1 | 1 | 0% | 2,114 | 2,087 | -1% | 0 | 0 | — |
case-07 | fail→pass | 4,649 | 4,029 | -13% | 1 | 1 | 0% | 527 | 1,701 | +223% | 0 | 0 | — |
case-08 | fail→pass | 38,991 | 21,133 | -46% | 1 | 1 | 0% | 2,001 | 2,061 | +3% | 0 | 0 | — |
case-09 | pass→pass | 11,164 | 10,079 | -10% | 1 | 1 | 0% | 1,523 | 1,590 | +4% | 0 | 0 | — |
case-10 | pass→pass | 7,491 | 2,731 | -64% | 1 | 1 | 0% | 1,018 | 1,504 | +48% | 0 | 0 | — |
case-11 | pass→pass | 13,071 | 11,754 | -10% | 1 | 1 | 0% | 1,780 | 1,796 | +1% | 0 | 0 | — |
case-12 | pass→fail | 20,791 | 32,468 | +56% | 1 | 1 | 0% | 1,437 | 1,620 | +13% | 0 | 0 | — |
case-13 | pass→pass | 11,538 | 13,311 | +15% | 1 | 1 | 0% | 1,437 | 1,920 | +34% | 0 | 0 | — |
case-14 | pass→pass | 7,132 | 8,109 | +14% | 1 | 1 | 0% | 1,031 | 2,388 | +132% | 0 | 0 | — |
case-15 | pass→pass | 9,555 | 3,269 | -66% | 1 | 1 | 0% | 1,608 | 1,596 | -1% | 0 | 0 | — |
case-16 | pass→pass | 4,007 | 2,917 | -27% | 1 | 1 | 0% | 423 | 1,426 | +237% | 0 | 0 | — |
case-17 | pass→pass | 8,271 | 4,003 | -52% | 1 | 1 | 0% | 1,089 | 1,463 | +34% | 0 | 0 | — |
case-18 | pass→pass | 9,932 | 4,402 | -56% | 1 | 1 | 0% | 1,513 | 1,599 | +6% | 0 | 0 | — |
case-19 | fail→pass | 36,448 | 15,436 | -58% | 1 | 1 | 0% | 1,165 | 1,972 | +69% | 0 | 0 | — |
case-20 | fail→pass | 14,460 | 7,946 | -45% | 1 | 1 | 0% | 2,154 | 1,859 | -14% | 0 | 0 | — |
case-21 | pass→pass | 28,430 | 7,089 | -75% | 1 | 1 | 0% | 1,801 | 1,866 | +4% | 0 | 0 | — |
case-22 | fail→pass | 13,053 | 9,081 | -30% | 1 | 1 | 0% | 2,162 | 2,614 | +21% | 0 | 0 | — |
case-23 | pass→pass | 11,162 | 3,092 | -72% | 1 | 1 | 0% | 1,586 | 1,555 | -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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 9/8/2026 | +36% |
| gemini-3.6-flash | verified | 8/7/2026 | +59% |
Other measured skills in the registry, with their headline benchmark lift.