Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Pushes all branches in a stack and opens or updates one dependent PR per slice. Use after stack-create to publish the stack or after adding commits to a slice.
.claude/skills/athola-stack-push/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 5% | 0% |
Push all branches in a stack and open (or update) one PR per slice, with each PR targeting its parent branch as base.
Run stack-push after stack-create has initialized the branch topology and at least one commit exists on each slice branch. Also run it after adding commits to any slice to update open PRs.
sanctum:stack-create)sanctum:stack-rebase)
stack-create completed: branches exist with commitsgh CLI authenticated (gh auth status)Create TodoWrite items before starting:
stack-push:branches-listedstack-push:branches-pushedstack-push:prs-openedstack-push:stack-summary-postedbranches-listed)Identify all branches in the stack. By convention they share a stack/<feature-name>/ prefix:
bashSTACK=stack/my-feature BASE=master git branch --list "${STACK}/*" | sed 's/^[* ]*//'
Verify each branch has commits beyond the base:
bashfor branch in $(git branch --list "${STACK}/*" \ | sed 's/^[* ]*//'); do count=$(git rev-list --count \ "$(git merge-base ${BASE} ${branch})..${branch}") echo "${branch}: ${count} commit(s)" done
Any branch showing 0 commits must have work added before pushing.
branches-pushed)Push every slice branch to the remote in stack order (base slice first):
bashfor branch in $(git branch --list "${STACK}/*" \ | sed 's/^[* ]*//' | sort); do git push --set-upstream origin "${branch}" echo "pushed: ${branch}" done
bashjj git push --all
prs-opened)Create one PR per slice, targeting its parent branch. For the first slice the parent is master; for subsequent slices the parent is the previous slice's branch.
bashPREV_BASE=master for branch in $(git branch --list "${STACK}/*" \ | sed 's/^[* ]*//' | sort); do # Check if a PR already exists for this branch existing=$(gh pr list \ --head "${branch}" \ --json number \ --jq '.[0].number' 2>/dev/null) if [ -n "${existing}" ]; then echo "PR #${existing} already open for ${branch} -- skipping" else gh pr create \ --base "${PREV_BASE}" \ --head "${branch}" \ --title "[$(echo ${branch} | sed 's|.*/||')] <title>" \ --body "Part of stack \`${STACK}\`." \ --draft echo "opened PR for ${branch} (base: ${PREV_BASE})" fi PREV_BASE="${branch}" done
Fill in the actual PR titles and bodies before removing the --draft flag. Run Skill(sanctum:pr-prep) on each slice to generate quality-gated descriptions.
stack-summary-posted)After all PRs are open, post a summary comment on the first PR (the root of the stack) listing the full chain:
bashROOT_PR=$(gh pr list \ --head "${STACK}/$(git branch --list \ "${STACK}/*" | sed 's/^[* ]*//' | sort | head -1 \ | sed 's|.*/||')" \ --json number --jq '.[0].number') # Build the stack table BODY="## Stack\n\n| # | Branch | PR |\n|---|--------|----|\n" N=1 for branch in $(git branch --list "${STACK}/*" \ | sed 's/^[* ]*//' | sort); do pr_num=$(gh pr list --head "${branch}" \ --json number --jq '.[0].number') BODY="${BODY}| ${N} | \`${branch}\` | #${pr_num} |\n" N=$((N+1)) done gh pr comment "${ROOT_PR}" --body "$(printf "${BODY}")"
branch before the child PR is opened
--draft by default; promote slices to readyindividually as they pass review
git rebase --update-refs run, re-push withgit push --force-with-lease (never --force)
Skill(sanctum:stack-rebase)to cascade the rebase before updating the next PR's base
stack-push:branches-listed throughstack-push:stack-summary-posted) are created before pushing starts and marked complete in order
before git push is called; branches with 0 commits are blocked
as base (root slice targets master/main)
## Stack comment containing a table withcolumns #, Branch, and PR listing all slices
gh pr list --head <branch> returns a PR number for eachpushed branch after the step completes
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 16,611 | 7,238 | -56% | 1 | 1 | 0% | 3,220 | 1,786 | -45% | 0 | 0 | — |
case-02 | fail→fail | 2,922 | 10,506 | +260% | 1 | 1 | 0% | 315 | 1,749 | +455% | 0 | 0 | — |
case-03 | fail→fail | 5,670 | 3,605 | -36% | 1 | 1 | 0% | 254 | 1,819 | +616% | 0 | 0 | — |
case-08 | pass→pass | 5,873 | 3,228 | -45% | 1 | 1 | 0% | 1,077 | 1,946 | +81% | 0 | 0 | — |
case-04 | fail→pass | 11,327 | 9,013 | -20% | 1 | 1 | 0% | 2,048 | 3,194 | +56% | 0 | 0 | — |
case-05 | pass→pass | 6,696 | 1,970 | -71% | 1 | 1 | 0% | 1,101 | 1,703 | +55% | 0 | 0 | — |
case-06 | pass→pass | 9,578 | 3,205 | -67% | 1 | 1 | 0% | 1,645 | 1,867 | +13% | 0 | 0 | — |
case-07 | pass→pass | 2,250 | 1,923 | -15% | 1 | 1 | 0% | 363 | 1,683 | +364% | 0 | 0 | — |
case-09 | fail→pass | 6,288 | 2,652 | -58% | 1 | 1 | 0% | 900 | 1,635 | +82% | 0 | 0 | — |
case-10 | fail→pass | 13,209 | 2,189 | -83% | 1 | 1 | 0% | 2,057 | 1,733 | -16% | 0 | 0 | — |
case-11 | pass→pass | 8,025 | 2,452 | -69% | 1 | 1 | 0% | 1,512 | 1,748 | +16% | 0 | 0 | — |
case-12 | pass→pass | 4,872 | 2,127 | -56% | 1 | 1 | 0% | 744 | 1,734 | +133% | 0 | 0 | — |
case-13 | pass→pass | 4,512 | 2,428 | -46% | 1 | 1 | 0% | 772 | 1,747 | +126% | 0 | 0 | — |
case-14 | pass→pass | 10,444 | 1,615 | -85% | 1 | 1 | 0% | 1,733 | 1,603 | -8% | 0 | 0 | — |
case-15 | fail→pass | 13,714 | 7,261 | -47% | 1 | 1 | 0% | 2,044 | 2,750 | +35% | 0 | 0 | — |
case-16 | pass→pass | 5,395 | 2,452 | -55% | 1 | 1 | 0% | 940 | 1,775 | +89% | 0 | 0 | — |
case-17 | fail→pass | 9,975 | 1,945 | -81% | 1 | 1 | 0% | 1,582 | 1,658 | +5% | 0 | 0 | — |
case-18 | fail→pass | 28,970 | 3,264 | -89% | 1 | 1 | 0% | 2,664 | 1,985 | -25% | 0 | 0 | — |
case-19 | pass→pass | 11,056 | 3,202 | -71% | 1 | 1 | 0% | 1,637 | 1,964 | +20% | 0 | 0 | — |
case-20 | fail→pass | 11,543 | 3,719 | -68% | 1 | 1 | 0% | 1,879 | 1,829 | -3% | 0 | 0 | — |
case-21 | fail→pass | 13,404 | 5,698 | -57% | 1 | 1 | 0% | 2,102 | 2,527 | +20% | 0 | 0 | — |
case-22 | fail→pass | 6,799 | 6,349 | -7% | 1 | 1 | 0% | 1,212 | 2,525 | +108% | 0 | 0 | — |
case-23 | fail→pass | 10,211 | 6,981 | -32% | 1 | 1 | 0% | 1,642 | 2,557 | +56% | 0 | 0 | — |
case-24 | pass→pass | 8,769 | 7,096 | -19% | 1 | 1 | 0% | 1,447 | 2,525 | +74% | 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. 24 cases were attempted, and 21 counted toward the lift figure. The other 3 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 +42 percentage points is the difference between those two pass rates over the 21 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.