Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Reviews MSBuild code changes using a 24-dimension methodology. Activates for code review, PR review, pull request analysis, design review, architecture review, code quality assessment, or style check of MSBuild code. Covers backwards compatibility, ChangeWave discipline, performance, allocation awareness, test coverage, error message quality, logging, string comparison, API surface, target authoring, cross-platform correctness, code simplification, concurrency, naming, SDK integration, evaluatio
.claude/skills/dotnet-reviewing-msbuild-code/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 66% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-03 | ✓→✗ | ▼ Worse | -81% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -42% | 0% |
Invoke @expert-reviewer for 24-dimension MSBuild code review. The agent contains the full methodology, principles, dimension rules with checklists, folder hotspot mapping, and 4-wave workflow.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 32,708 | 7,752 | -76% | 1 | 1 | 0% | 6,326 | 871 | -86% | 0 | 0 | — |
case-02 | pass→pass | 25,710 | 26,946 | +5% | 1 | 1 | 0% | 4,285 | 5,113 | +19% | 0 | 0 | — |
case-03 | pass→fail | 40,159 | 7,836 | -80% | 1 | 1 | 0% | 6,168 | 1,176 | -81% | 0 | 0 | — |
case-04 | pass→pass | 14,924 | 18,623 | +25% | 1 | 1 | 0% | 3,144 | 3,874 | +23% | 0 | 0 | — |
case-05 | pass→pass | 20,640 | 23,071 | +12% | 1 | 1 | 0% | 2,891 | 3,253 | +13% | 0 | 0 | — |
case-06 | pass→pass | 13,791 | 9,534 | -31% | 1 | 1 | 0% | 2,306 | 1,937 | -16% | 0 | 0 | — |
case-07 | fail→fail | 16,160 | 4,884 | -70% | 1 | 1 | 0% | 2,142 | 769 | -64% | 0 | 0 | — |
case-08 | pass→fail | 29,103 | 19,176 | -34% | 1 | 1 | 0% | 5,389 | 3,129 | -42% | 0 | 0 | — |
case-09 | fail→pass | 33,007 | 31,283 | -5% | 1 | 1 | 0% | 4,718 | 4,588 | -3% | 0 | 0 | — |
case-10 | fail→pass | 27,030 | 41,336 | +53% | 1 | 1 | 0% | 3,799 | 6,314 | +66% | 0 | 0 | — |
case-11 | pass→pass | 24,200 | 25,038 | +3% | 1 | 1 | 0% | 3,718 | 4,063 | +9% | 0 | 0 | — |
case-12 | pass→pass | 28,780 | 39,633 | +38% | 1 | 1 | 0% | 4,375 | 6,144 | +40% | 0 | 0 | — |
case-13 | pass→pass | 17,516 | 13,941 | -20% | 1 | 1 | 0% | 2,743 | 2,244 | -18% | 0 | 0 | — |
case-14 | fail→fail | 10,678 | 12,626 | +18% | 1 | 1 | 0% | 1,483 | 1,740 | +17% | 0 | 0 | — |
case-15 | fail→fail | 53,847 | 32,557 | -40% | 1 | 1 | 0% | 8,226 | 5,836 | -29% | 0 | 0 | — |
case-16 | fail→fail | 12,159 | 10,562 | -13% | 1 | 1 | 0% | 1,976 | 1,537 | -22% | 0 | 0 | — |
case-17 | fail→fail | 25,504 | 35,194 | +38% | 1 | 1 | 0% | 4,368 | 6,363 | +46% | 0 | 0 | — |
case-18 | pass→pass | 24,697 | 30,542 | +24% | 1 | 1 | 0% | 4,257 | 5,951 | +40% | 0 | 0 | — |
case-19 | fail→pass | 44,614 | 46,213 | +4% | 1 | 1 | 0% | 6,875 | 6,825 | -1% | 0 | 0 | — |
case-20 | pass→fail | 25,023 | 11,503 | -54% | 1 | 1 | 0% | 3,647 | 1,801 | -51% | 0 | 0 | — |
case-21 | pass→fail | 18,708 | 14,510 | -22% | 1 | 1 | 0% | 3,215 | 2,360 | -27% | 0 | 0 | — |
case-22 | pass→pass | 38,200 | 36,472 | -5% | 1 | 1 | 0% | 6,671 | 6,465 | -3% | 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 -14 percentage points is the difference between those two pass rates over the 22 comparable cases. 4 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.