Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when recording TiDB integration tests under tests/integrationtest and verifying regenerated result files stay minimal and correct.
.claude/skills/pingcap-tidb-integrationtest-recorder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -77% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -60% | 0% |
Use this workflow for changes under tests/integrationtest/t/** or when SQL behavior needs integration coverage. Do not use -record for this suite. Canonical command details live in docs/agents/testing-flow.md -> Integration tests (/tests/integrationtest).
docs/agents/testing-flow.md -> Integration tests (/tests/integrationtest) for the recording command.TestName from the path under tests/integrationtest/t/ without the .test suffix (example: planner/core/binary_plan).tests/integrationtest/r/** and keep result diffs minimal.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,474 | 8,859 | -49% | 1 | 1 | 0% | 2,779 | 1,578 | -43% | 0 | 0 | — |
case-02 | fail→pass | 13,235 | 8,663 | -35% | 1 | 1 | 0% | 1,992 | 1,625 | -18% | 0 | 0 | — |
case-03 | fail→pass | 30,539 | 5,215 | -83% | 1 | 1 | 0% | 2,633 | 1,180 | -55% | 0 | 0 | — |
case-04 | pass→pass | 10,654 | 4,125 | -61% | 1 | 1 | 0% | 1,737 | 793 | -54% | 0 | 0 | — |
case-05 | pass→pass | 21,621 | 4,133 | -81% | 1 | 1 | 0% | 1,689 | 666 | -61% | 0 | 0 | — |
case-06 | pass→pass | 12,833 | 3,204 | -75% | 1 | 1 | 0% | 2,200 | 727 | -67% | 0 | 0 | — |
case-07 | fail→pass | 12,983 | 3,811 | -71% | 1 | 1 | 0% | 1,751 | 769 | -56% | 0 | 0 | — |
case-08 | fail→pass | 13,191 | 3,436 | -74% | 1 | 1 | 0% | 1,914 | 434 | -77% | 0 | 0 | — |
case-09 | fail→pass | 9,383 | 2,908 | -69% | 1 | 1 | 0% | 1,428 | 577 | -60% | 0 | 0 | — |
case-10 | pass→pass | 15,564 | 2,890 | -81% | 1 | 1 | 0% | 2,145 | 573 | -73% | 0 | 0 | — |
case-11 | pass→pass | 17,004 | 8,030 | -53% | 1 | 1 | 0% | 2,304 | 1,396 | -39% | 0 | 0 | — |
case-12 | pass→pass | 45,225 | 5,130 | -89% | 1 | 1 | 0% | 1,881 | 970 | -48% | 0 | 0 | — |
case-13 | pass→pass | 12,493 | 3,233 | -74% | 1 | 1 | 0% | 1,772 | 652 | -63% | 0 | 0 | — |
case-14 | fail→pass | 5,356 | 2,759 | -48% | 1 | 1 | 0% | 698 | 484 | -31% | 0 | 0 | — |
case-15 | pass→pass | 14,987 | 6,032 | -60% | 1 | 1 | 0% | 2,262 | 1,023 | -55% | 0 | 0 | — |
case-16 | pass→pass | 12,961 | 9,970 | -23% | 1 | 1 | 0% | 2,121 | 1,184 | -44% | 0 | 0 | — |
case-17 | pass→pass | 10,492 | 4,099 | -61% | 1 | 1 | 0% | 1,499 | 796 | -47% | 0 | 0 | — |
case-18 | fail→pass | 9,887 | 9,222 | -7% | 1 | 1 | 0% | 1,596 | 637 | -60% | 0 | 0 | — |
case-19 | pass→pass | 17,226 | 4,961 | -71% | 1 | 1 | 0% | 2,311 | 978 | -58% | 0 | 0 | — |
case-20 | fail→pass | 13,132 | 4,906 | -63% | 1 | 1 | 0% | 1,975 | 904 | -54% | 0 | 0 | — |
case-21 | fail→pass | 6,529 | 3,053 | -53% | 1 | 1 | 0% | 639 | 579 | -9% | 0 | 0 | — |
case-22 | pass→pass | 10,029 | 4,252 | -58% | 1 | 1 | 0% | 1,475 | 759 | -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. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 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.