Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when deciding whether make bazel_prepare is required before build or test commands based on local file changes in TiDB.
.claude/skills/pingcap-tidb-bazel-prepare-gate/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -47% | 0% |
Use this skill before build/test commands when you are unsure whether make bazel_prepare is required. Policy source: AGENTS.md -> Build Flow -> When make bazel_prepare is required. In normal coding loops, skip this skill unless one of the decision-rule triggers is likely present.
Run from repository root:
bashgit status --short git diff --name-status git diff --name-status --cached git ls-files --others --exclude-standard git diff -U0 -- '*.go' git diff -U0 --cached -- '*.go'
Trigger conditions are defined in AGENTS.md -> Build Flow -> When make bazel_prepare is required. Compare the output from the commands above against those conditions.
For the top-level test-function trigger in existing *_test.go files, inspect added lines in git diff -U0 -- '*.go' and git diff -U0 --cached -- '*.go' output for patterns like:
diff+func TestXxx(t *testing.T) {
and treat that as requiring make bazel_prepare.
For import-section changes in existing Go files, inspect git diff -U0 -- '*.go' and git diff -U0 --cached -- '*.go' for added/removed import lines, for example:
diff+import ( -import "fmt" + "context"
and treat those changes as requiring make bazel_prepare.
If any condition matches, run make bazel_prepare. If none of the rules match, continue without make bazel_prepare and report the evidence.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,959 | 34,587 | +189% | 1 | 1 | 0% | 703 | 659 | -6% | 0 | 0 | — |
case-02 | fail→fail | 7,212 | 6,031 | -16% | 1 | 1 | 0% | 330 | 658 | +99% | 0 | 0 | — |
case-03 | fail→fail | 11,328 | 10,423 | -8% | 1 | 1 | 0% | 519 | 602 | +16% | 0 | 0 | — |
case-04 | fail→pass | 11,104 | 13,695 | +23% | 1 | 1 | 0% | 1,842 | 1,038 | -44% | 0 | 0 | — |
case-05 | fail→pass | 10,674 | 8,843 | -17% | 1 | 1 | 0% | 1,029 | 1,026 | -0% | 0 | 0 | — |
case-06 | pass→pass | 7,641 | 14,148 | +85% | 1 | 1 | 0% | 1,158 | 1,107 | -4% | 0 | 0 | — |
case-07 | pass→pass | 12,912 | 3,209 | -75% | 1 | 1 | 0% | 1,039 | 800 | -23% | 0 | 0 | — |
case-08 | pass→pass | 3,804 | 2,445 | -36% | 1 | 1 | 0% | 514 | 670 | +30% | 0 | 0 | — |
case-09 | fail→pass | 6,928 | 2,990 | -57% | 1 | 1 | 0% | 870 | 785 | -10% | 0 | 0 | — |
case-10 | pass→pass | 7,891 | 2,819 | -64% | 1 | 1 | 0% | 1,197 | 819 | -32% | 0 | 0 | — |
case-11 | fail→fail | 12,684 | 6,875 | -46% | 1 | 1 | 0% | 1,928 | 726 | -62% | 0 | 0 | — |
case-12 | pass→fail | 10,065 | 7,643 | -24% | 1 | 1 | 0% | 998 | 844 | -15% | 0 | 0 | — |
case-13 | pass→pass | 4,539 | 3,683 | -19% | 1 | 1 | 0% | 461 | 865 | +88% | 0 | 0 | — |
case-14 | pass→fail | 13,238 | 7,096 | -46% | 1 | 1 | 0% | 2,060 | 1,478 | -28% | 0 | 0 | — |
case-15 | pass→pass | 4,391 | 9,053 | +106% | 1 | 1 | 0% | 679 | 891 | +31% | 0 | 0 | — |
case-16 | pass→pass | 4,469 | 3,048 | -32% | 1 | 1 | 0% | 345 | 599 | +74% | 0 | 0 | — |
case-17 | pass→pass | 6,449 | 3,376 | -48% | 1 | 1 | 0% | 797 | 773 | -3% | 0 | 0 | — |
case-18 | fail→pass | 9,288 | 2,931 | -68% | 1 | 1 | 0% | 1,430 | 786 | -45% | 0 | 0 | — |
case-19 | fail→pass | 10,954 | 3,228 | -71% | 1 | 1 | 0% | 1,555 | 822 | -47% | 0 | 0 | — |
case-20 | fail→pass | 14,411 | 4,073 | -72% | 1 | 1 | 0% | 1,308 | 932 | -29% | 0 | 0 | — |
case-21 | pass→pass | 10,603 | 14,090 | +33% | 1 | 1 | 0% | 1,162 | 1,091 | -6% | 0 | 0 | — |
case-22 | fail→fail | 9,100 | 10,480 | +15% | 1 | 1 | 0% | 1,475 | 758 | -49% | 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 16 counted toward the lift figure. The other 6 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 +18 percentage points is the difference between those two pass rates over the 16 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.