Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →SQL database migrations with zero-downtime strategies for PostgreSQL, MySQL, SQL Server
.claude/skills/dokhacgiakhoa-database-migrations-sql-migrations/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 77% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-10 | ✓→✗ | ▼ Worse | 83% | 0% |
You are a SQL database migration expert specializing in zero-downtime deployments, data integrity, and production-ready migration strategies for PostgreSQL, MySQL, and SQL Server. Create comprehensive migration scripts with rollback procedures, validation checks, and performance optimization.
The user needs SQL database migrations that ensure data integrity, minimize downtime, and provide safe rollback options. Focus on production-ready strategies that handle edge cases, large datasets, and concurrent operations.
$ARGUMENTS
resources/implementation-playbook.md.Focus on production-ready SQL migrations with zero-downtime deployment strategies, comprehensive validation, and enterprise-grade safety mechanisms.
resources/implementation-playbook.md for detailed patterns and examples.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 49,410 | 60,346 | +22% | 1 | 1 | 0% | 8,264 | 8,673 | +5% | 0 | 0 | — |
case-02 | fail→fail | 27,203 | 60,512 | +122% | 1 | 1 | 0% | 4,232 | 4,690 | +11% | 0 | 0 | — |
case-03 | pass→pass | 35,340 | 31,821 | -10% | 1 | 1 | 0% | 6,410 | 6,563 | +2% | 0 | 0 | — |
case-04 | fail→pass | 39,204 | 32,190 | -18% | 1 | 1 | 0% | 5,811 | 6,677 | +15% | 0 | 0 | — |
case-05 | fail→pass | 15,828 | 71,164 | +350% | 1 | 1 | 0% | 3,082 | 5,446 | +77% | 0 | 0 | — |
case-06 | pass→pass | 23,917 | 28,015 | +17% | 1 | 1 | 0% | 3,834 | 5,028 | +31% | 0 | 0 | — |
case-07 | fail→pass | 30,589 | 49,449 | +62% | 1 | 1 | 0% | 5,555 | 8,122 | +46% | 0 | 0 | — |
case-08 | pass→pass | 15,717 | 21,094 | +34% | 1 | 1 | 0% | 2,366 | 4,051 | +71% | 0 | 0 | — |
case-09 | fail→fail | 26,310 | 37,742 | +43% | 1 | 1 | 0% | 3,741 | 7,456 | +99% | 0 | 0 | — |
case-10 | pass→fail | 22,554 | 37,280 | +65% | 1 | 1 | 0% | 3,998 | 7,311 | +83% | 0 | 0 | — |
case-11 | pass→pass | 36,712 | 31,173 | -15% | 1 | 1 | 0% | 7,151 | 6,318 | -12% | 0 | 0 | — |
case-12 | pass→pass | 27,265 | 35,648 | +31% | 1 | 1 | 0% | 4,291 | 7,306 | +70% | 0 | 0 | — |
case-13 | fail→pass | 24,247 | 36,538 | +51% | 1 | 1 | 0% | 4,041 | 7,366 | +82% | 0 | 0 | — |
case-14 | fail→fail | 21,055 | 35,036 | +66% | 1 | 1 | 0% | 3,560 | 7,237 | +103% | 0 | 0 | — |
case-15 | pass→pass | 18,326 | 26,691 | +46% | 1 | 1 | 0% | 3,502 | 5,402 | +54% | 0 | 0 | — |
case-16 | pass→pass | 28,019 | 47,786 | +71% | 1 | 1 | 0% | 4,757 | 7,836 | +65% | 0 | 0 | — |
case-17 | pass→pass | 20,385 | 23,766 | +17% | 1 | 1 | 0% | 3,832 | 5,394 | +41% | 0 | 0 | — |
case-18 | fail→fail | 28,289 | 23,155 | -18% | 1 | 1 | 0% | 5,579 | 4,466 | -20% | 0 | 0 | — |
case-19 | pass→fail | 25,643 | 30,703 | +20% | 1 | 1 | 0% | 5,145 | 5,814 | +13% | 0 | 0 | — |
case-20 | pass→pass | 18,032 | 33,417 | +85% | 1 | 1 | 0% | 3,723 | 6,089 | +64% | 0 | 0 | — |
case-21 | pass→pass | 21,563 | 25,544 | +18% | 1 | 1 | 0% | 3,385 | 4,293 | +27% | 0 | 0 | — |
case-22 | pass→pass | 14,155 | 24,993 | +77% | 1 | 1 | 0% | 2,692 | 5,336 | +98% | 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 +9 percentage points is the difference between those two pass rates over the 22 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.