Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate a new pkg/linters analyzer from a merged or open PR pattern.
.claude/skills/github-pr-to-go-linter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-10 | ✓→✗ | ▼ Worse | 21% | 0% |
| case-22 | ✓→✗ | ▼ Worse | -88% | 0% |
Use this skill when a user asks to create a new custom Go linter based on a code pattern fixed in a pull request.
Convert one concrete PR pattern into a new go/analysis linter under pkg/linters/<name>/ with tests and runner registration.
pkg/linters/ already covers it.pkg/linters/<name>/<name>.go with exported Analyzerpkg/linters/<name>/<name>_test.go using analysistestpkg/linters/<name>/testdata/src/<name>/<name>.go fixtures with // wantcmd/linters/main.go registration in multichecker.Main(...)go test ./pkg/linters/<name>/...go build ./cmd/lintersmake golint-customFor PR #33038 (Refactor pkg mutex sites to use deferred unlocks consistently), derive a linter idea that reports lock/unlock sections that manually unlock instead of deferring unlock immediately after lock when the function body matches the same cache/logger-style critical section pattern.
pkg/linters/<name>/ and cmd/linters/main.go.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,098 | 7,111 | +17% | 1 | 1 | 0% | 339 | 997 | +194% | 0 | 0 | — |
case-02 | fail→fail | 6,138 | 16,944 | +176% | 1 | 1 | 0% | 261 | 3,907 | +1397% | 0 | 0 | — |
case-03 | fail→fail | 19,522 | 5,448 | -72% | 1 | 1 | 0% | 3,939 | 708 | -82% | 0 | 0 | — |
case-04 | fail→pass | 27,581 | 14,885 | -46% | 1 | 1 | 0% | 3,975 | 3,690 | -7% | 0 | 0 | — |
case-05 | fail→fail | 22,283 | 32,508 | +46% | 1 | 1 | 0% | 5,047 | 6,707 | +33% | 0 | 0 | — |
case-06 | pass→pass | 9,408 | 3,482 | -63% | 1 | 1 | 0% | 1,586 | 1,172 | -26% | 0 | 0 | — |
case-07 | fail→fail | 6,667 | 2,499 | -63% | 1 | 1 | 0% | 949 | 880 | -7% | 0 | 0 | — |
case-08 | pass→pass | 13,652 | 5,874 | -57% | 1 | 1 | 0% | 2,075 | 1,327 | -36% | 0 | 0 | — |
case-09 | fail→pass | 10,374 | 5,213 | -50% | 1 | 1 | 0% | 1,717 | 1,400 | -18% | 0 | 0 | — |
case-10 | pass→fail | 5,005 | 3,893 | -22% | 1 | 1 | 0% | 813 | 986 | +21% | 0 | 0 | — |
case-11 | pass→pass | 8,077 | 6,402 | -21% | 1 | 1 | 0% | 1,451 | 1,571 | +8% | 0 | 0 | — |
case-12 | fail→fail | 12,828 | 5,523 | -57% | 1 | 1 | 0% | 2,145 | 1,484 | -31% | 0 | 0 | — |
case-13 | pass→pass | 7,390 | 3,086 | -58% | 1 | 1 | 0% | 1,328 | 994 | -25% | 0 | 0 | — |
case-14 | pass→pass | 5,194 | 2,912 | -44% | 1 | 1 | 0% | 771 | 764 | -1% | 0 | 0 | — |
case-15 | fail→pass | 7,014 | 2,357 | -66% | 1 | 1 | 0% | 1,270 | 906 | -29% | 0 | 0 | — |
case-16 | pass→pass | 4,007 | 3,132 | -22% | 1 | 1 | 0% | 614 | 1,080 | +76% | 0 | 0 | — |
case-17 | fail→fail | 6,650 | 5,162 | -22% | 1 | 1 | 0% | 1,105 | 1,252 | +13% | 0 | 0 | — |
case-18 | pass→pass | 7,148 | 4,104 | -43% | 1 | 1 | 0% | 1,084 | 1,147 | +6% | 0 | 0 | — |
case-19 | pass→pass | 7,687 | 18,619 | +142% | 1 | 1 | 0% | 1,319 | 4,756 | +261% | 0 | 0 | — |
case-20 | pass→pass | 13,068 | 11,415 | -13% | 1 | 1 | 0% | 1,124 | 1,514 | +35% | 0 | 0 | — |
case-21 | pass→pass | 5,933 | 8,274 | +39% | 1 | 1 | 0% | 1,094 | 1,804 | +65% | 0 | 0 | — |
case-22 | pass→fail | 33,884 | 6,562 | -81% | 1 | 1 | 0% | 6,196 | 766 | -88% | 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 18 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 +5 percentage points is the difference between those two pass rates over the 18 comparable cases. 3 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.