Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create a git branch following Sentry naming conventions. Use when asked to "create a branch", "new branch", "start a branch", "make a branch", "switch to a new branch", or when starting new work on the default branch.
.claude/skills/sickn33-create-branch/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -9% | 0% |
| case-16 | ✓→✗ | ▼ Worse | 125% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 57% | 0% |
| case-09 | ✓→✓ | = Same ✓ | 120% | 0% |
Create a git branch with the correct type prefix and a descriptive name following Sentry conventions.
feat, fix, docs, or another branch type.Run gh api user --jq .login to get the GitHub username.
If the command fails (e.g. not authenticated), ask the user for their preferred prefix.
If $ARGUMENTS is provided, use it as the description of the work.
If no arguments, check for local changes:
bashgit diff git diff --cached git status --short
Pick the type from this table based on the description:
| Type | Use when | | --------- | --------------------------------------------------------------------- | | feat | New user-facing functionality | | fix | Broken behavior now works | | ref | Same behavior, different structure | | chore | Deps, config, version bumps, updating existing tooling — no new logic | | perf | Same behavior, faster | | style | CSS, formatting, visual-only | | docs | Documentation only | | test | Tests only | | ci | CI/CD config | | build | Build system | | meta | Repo metadata changes | | license | License changes |
When unsure: feat for new things (including new scripts, skills, or tools), ref for restructuring existing things, chore only when updating/maintaining something that already exists.
Build the branch name as <username>/<type>/<short-description>.
Rules for <short-description>:
Present it to the user and ask if they want to use it, modify it, or change the type.
| Work description | Branch name | | ------------------------------------------ | ------------------------------------------- | | Dropdown menu not closing on outside click | priscila/fix/dropdown-not-closing-on-blur | | Adding search to conversations page | priscila/feat/add-search-to-conversations | | Restructuring drawer components | priscila/ref/simplify-drawer-components | | Updating test fixtures | priscila/chore/update-test-fixtures | | Bumping @sentry/react to latest version | priscila/chore/bump-sentry-react | | Adding a new agent skill | priscila/feat/add-create-branch-skill |
Once confirmed, detect the current and default branch:
bashgit branch --show-current git remote | grep -qx origin && echo origin || git remote | head -1 git symbolic-ref refs/remotes/<remote>/HEAD 2>/dev/null | sed 's|refs/remotes/<remote>/||' | tr -d '[:space:]'
If symbolic-ref fails, fall back to git branch --list main master: use the one that exists; if both or neither exist, ask the user.
If git branch --show-current is empty (detached HEAD), show the current commit (git rev-parse --short HEAD) and ask whether to branch from it or switch to the default branch first.
Otherwise, if the current branch is not the default branch, warn the user and ask whether to branch from the current branch or switch to the default branch first.
If the user wants to switch to the default branch, handle any uncommitted changes appropriately (offer to stash them if present), then run git checkout <default-branch>. On any failure, restore stashed changes if applicable and stop.
Before creating the branch, check that the name doesn't already exist locally or on the remote (git show-ref). If it does, ask the user to choose a different name.
Create the branch:
bashgit checkout -b <branch-name>
Restore any stashed changes after the branch is created.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,664 | 4,689 | -30% | 1 | 1 | 0% | 1,077 | 1,468 | +36% | 0 | 0 | — |
case-02 | fail→fail | 9,702 | 4,110 | -58% | 1 | 1 | 0% | 1,078 | 1,561 | +45% | 0 | 0 | — |
case-03 | fail→fail | 6,811 | 3,560 | -48% | 1 | 1 | 0% | 507 | 1,361 | +168% | 0 | 0 | — |
case-04 | fail→fail | 7,753 | 6,598 | -15% | 1 | 1 | 0% | 1,294 | 1,577 | +22% | 0 | 0 | — |
case-05 | fail→fail | 3,547 | 9,772 | +176% | 1 | 1 | 0% | 565 | 1,910 | +238% | 0 | 0 | — |
case-06 | fail→fail | 6,323 | 5,139 | -19% | 1 | 1 | 0% | 1,059 | 2,081 | +97% | 0 | 0 | — |
case-07 | pass→pass | 7,599 | 3,875 | -49% | 1 | 1 | 0% | 1,220 | 1,919 | +57% | 0 | 0 | — |
case-08 | fail→fail | 4,978 | 6,180 | +24% | 1 | 1 | 0% | 848 | 2,282 | +169% | 0 | 0 | — |
case-09 | pass→pass | 5,841 | 5,988 | +3% | 1 | 1 | 0% | 1,081 | 2,373 | +120% | 0 | 0 | — |
case-10 | fail→fail | 8,299 | 5,150 | -38% | 1 | 1 | 0% | 1,399 | 2,212 | +58% | 0 | 0 | — |
case-11 | fail→fail | 7,657 | 6,489 | -15% | 1 | 1 | 0% | 1,330 | 1,599 | +20% | 0 | 0 | — |
case-12 | pass→fail | 10,540 | 8,093 | -23% | 1 | 1 | 0% | 1,986 | 1,811 | -9% | 0 | 0 | — |
case-13 | fail→fail | 7,866 | 5,038 | -36% | 1 | 1 | 0% | 1,395 | 2,097 | +50% | 0 | 0 | — |
case-14 | fail→pass | 12,262 | 5,985 | -51% | 1 | 1 | 0% | 2,158 | 2,249 | +4% | 0 | 0 | — |
case-15 | pass→pass | 5,626 | 8,846 | +57% | 1 | 1 | 0% | 927 | 1,947 | +110% | 0 | 0 | — |
case-16 | pass→fail | 5,408 | 8,110 | +50% | 1 | 1 | 0% | 808 | 1,821 | +125% | 0 | 0 | — |
case-17 | pass→pass | 3,653 | 2,821 | -23% | 1 | 1 | 0% | 595 | 1,719 | +189% | 0 | 0 | — |
case-18 | fail→fail | 8,640 | 7,132 | -17% | 1 | 1 | 0% | 1,370 | 1,710 | +25% | 0 | 0 | — |
case-19 | pass→pass | 7,503 | 5,012 | -33% | 1 | 1 | 0% | 1,243 | 2,058 | +66% | 0 | 0 | — |
case-20 | pass→pass | 5,814 | 8,120 | +40% | 1 | 1 | 0% | 1,069 | 2,550 | +139% | 0 | 0 | — |
case-21 | pass→pass | 3,855 | 5,833 | +51% | 1 | 1 | 0% | 674 | 2,181 | +224% | 0 | 0 | — |
case-22 | pass→pass | 3,015 | 6,244 | +107% | 1 | 1 | 0% | 475 | 2,429 | +411% | 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. 22 cases were attempted, and 13 counted toward the lift figure. The other 9 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 -5 percentage points is the difference between those two pass rates over the 13 comparable cases. 2 cases got worse with the skill loaded, and they are 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.