Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Commit session changes to git. Identify files from session context, ask user pick message, run validation only if needed. Use when user want commit / save / land changes.
.claude/skills/platformplatform-commit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 55% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 115% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -15% | 0% |
Asked commit now != permission for future commits. Each commit needs explicit user instruction. No amend, revert, push without explicit ask.
Speed critical. No slow commands unless changes warrant.
git status / log / diff.Frontend changes touching user-facing strings: inspect .po files for missing translations and translate them. Missing keys: flag user; don't commit until resolved.
Skip if build / test / lint / format / aspire-restart / e2e already ran clean this session on same files. No re-run, no offer.
When need:
Run build first:
build (no flags - covers backend AND frontend; backend can break frontend via API contracts)build --frontendbuild --cliAfter build, parallel (--no-build where applicable):
format --<target>, lint --<target> - per target whose code changedtest --no-build - backend changeaspire-restart - service / Aspire wiring change (large only)e2e - large frontend or shared E2E-covered change; --smoke for low-impact (rare pre-commit)Judge target by file (project / manifest / lock files affect owning build, even when extension non-obvious).
Simple, focused commit: propose one message, commit.
Complex (multiple unrelated changes, doesn't fit one line, validation choices needed, file set ambiguous): use AskUserQuestion - 2-4 message options + conditionals in one call. No back-and-forth. Follow up only on "Other" or unexpected input.
No conventional commit prefixes (feat:, fix:, docs:, chore:).
Conditional questions:
Stage explicit files. Never git add -A or git add ..
Asked commit now != permission for future commits.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 2,335 | 3,000 | +28% | 1 | 1 | 0% | 308 | 644 | +109% | 0 | 0 | — |
case-02 | fail→fail | 8,386 | 3,285 | -61% | 1 | 1 | 0% | 1,388 | 755 | -46% | 0 | 0 | — |
case-03 | fail→fail | 4,764 | 5,208 | +9% | 1 | 1 | 0% | 221 | 705 | +219% | 0 | 0 | — |
case-04 | pass→pass | 12,997 | 7,445 | -43% | 1 | 1 | 0% | 2,069 | 1,767 | -15% | 0 | 0 | — |
case-05 | fail→fail | 9,368 | 5,170 | -45% | 1 | 1 | 0% | 1,626 | 1,284 | -21% | 0 | 0 | — |
case-06 | pass→pass | 6,779 | 3,718 | -45% | 1 | 1 | 0% | 1,157 | 1,063 | -8% | 0 | 0 | — |
case-07 | pass→pass | 8,199 | 3,062 | -63% | 1 | 1 | 0% | 1,260 | 1,015 | -19% | 0 | 0 | — |
case-08 | fail→pass | 6,414 | 1,873 | -71% | 1 | 1 | 0% | 1,093 | 760 | -30% | 0 | 0 | — |
case-09 | pass→pass | 4,668 | 2,665 | -43% | 1 | 1 | 0% | 803 | 984 | +23% | 0 | 0 | — |
case-10 | fail→pass | 3,712 | 2,244 | -40% | 1 | 1 | 0% | 545 | 843 | +55% | 0 | 0 | — |
case-11 | pass→pass | 8,399 | 1,864 | -78% | 1 | 1 | 0% | 1,323 | 755 | -43% | 0 | 0 | — |
case-12 | fail→pass | 3,311 | 2,499 | -25% | 1 | 1 | 0% | 412 | 885 | +115% | 0 | 0 | — |
case-13 | fail→fail | 8,898 | 1,952 | -78% | 1 | 1 | 0% | 1,373 | 746 | -46% | 0 | 0 | — |
case-14 | pass→pass | 12,464 | 3,128 | -75% | 1 | 1 | 0% | 1,981 | 999 | -50% | 0 | 0 | — |
case-15 | fail→fail | 11,858 | 4,594 | -61% | 1 | 1 | 0% | 1,765 | 1,295 | -27% | 0 | 0 | — |
case-20 | fail→pass | 12,050 | 1,430 | -88% | 1 | 1 | 0% | 1,784 | 670 | -62% | 0 | 0 | — |
case-16 | fail→pass | 6,341 | 2,222 | -65% | 1 | 1 | 0% | 982 | 835 | -15% | 0 | 0 | — |
case-17 | fail→pass | 9,030 | 2,066 | -77% | 1 | 1 | 0% | 1,351 | 791 | -41% | 0 | 0 | — |
case-18 | fail→pass | 9,519 | 1,586 | -83% | 1 | 1 | 0% | 1,606 | 712 | -56% | 0 | 0 | — |
case-19 | fail→pass | 9,390 | 1,912 | -80% | 1 | 1 | 0% | 1,482 | 773 | -48% | 0 | 0 | — |
case-21 | fail→pass | 7,086 | 1,752 | -75% | 1 | 1 | 0% | 1,004 | 758 | -25% | 0 | 0 | — |
case-22 | pass→pass | 4,786 | 2,009 | -58% | 1 | 1 | 0% | 681 | 759 | +11% | 0 | 0 | — |
case-23 | fail→pass | 10,619 | 1,499 | -86% | 1 | 1 | 0% | 1,809 | 701 | -61% | 0 | 0 | — |
case-24 | pass→pass | 10,597 | 1,929 | -82% | 1 | 1 | 0% | 1,770 | 726 | -59% | 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.
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.