Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Perform a structured code review using change detection and impact
.claude/skills/tirth8205-review-changes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -73% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -41% | 0% |
Review a change set with risk scores and blast radius from the knowledge graph.
detect_changes_tool for risk-scored changed functions, test gaps and affected flows.get_affected_flows_tool only when you need the steps of an affected flow.query_graph_tool with pattern="tests_for" to check test coverage.get_impact_radius_tool when the blast radius is not clear from step 1.Group findings by risk level (high, medium, low). For each finding give what changed and why it matters, its test coverage, and the suggested fix. End with a merge recommendation.
get_minimal_context_tool(task="<your task>") before any other graph tool.detail_level="minimal" wherever a tool accepts it. Use "standard" only when minimal is not enough.query_graph_tool call over a broad listing call.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 23,075 | 11,501 | -50% | 1 | 1 | 0% | 2,720 | 611 | -78% | 0 | 0 | — |
case-01 | fail→fail | 29,530 | 17,935 | -39% | 1 | 1 | 0% | 3,517 | 650 | -82% | 0 | 0 | — |
case-02 | fail→fail | 26,685 | 37,911 | +42% | 1 | 1 | 0% | 2,605 | 654 | -75% | 0 | 0 | — |
case-04 | fail→pass | 12,934 | 9,687 | -25% | 1 | 1 | 0% | 1,045 | 738 | -29% | 0 | 0 | — |
case-05 | fail→pass | 41,073 | 25,058 | -39% | 1 | 1 | 0% | 3,932 | 993 | -75% | 0 | 0 | — |
case-06 | fail→pass | 25,403 | 5,907 | -77% | 1 | 1 | 0% | 3,351 | 890 | -73% | 0 | 0 | — |
case-07 | fail→pass | 14,864 | 4,376 | -71% | 1 | 1 | 0% | 1,261 | 814 | -35% | 0 | 0 | — |
case-08 | fail→pass | 13,276 | 10,686 | -20% | 1 | 1 | 0% | 1,877 | 1,111 | -41% | 0 | 0 | — |
case-09 | fail→pass | 12,998 | 3,682 | -72% | 1 | 1 | 0% | 890 | 649 | -27% | 0 | 0 | — |
case-10 | pass→pass | 13,082 | 3,889 | -70% | 1 | 1 | 0% | 1,026 | 739 | -28% | 0 | 0 | — |
case-11 | pass→pass | 6,530 | 4,214 | -35% | 1 | 1 | 0% | 783 | 653 | -17% | 0 | 0 | — |
case-12 | fail→pass | 20,041 | 7,430 | -63% | 1 | 1 | 0% | 1,753 | 493 | -72% | 0 | 0 | — |
case-13 | fail→pass | 15,159 | 2,503 | -83% | 1 | 1 | 0% | 1,351 | 585 | -57% | 0 | 0 | — |
case-14 | fail→pass | 15,891 | 8,240 | -48% | 1 | 1 | 0% | 1,599 | 635 | -60% | 0 | 0 | — |
case-15 | fail→fail | 17,080 | 9,262 | -46% | 1 | 1 | 0% | 1,678 | 802 | -52% | 0 | 0 | — |
case-16 | pass→fail | 21,378 | 6,378 | -70% | 1 | 1 | 0% | 2,271 | 689 | -70% | 0 | 0 | — |
case-17 | pass→pass | 11,852 | 22,098 | +86% | 1 | 1 | 0% | 1,714 | 2,401 | +40% | 0 | 0 | — |
case-18 | pass→pass | 13,307 | 12,141 | -9% | 1 | 1 | 0% | 1,178 | 1,266 | +7% | 0 | 0 | — |
case-19 | fail→pass | 9,275 | 10,758 | +16% | 1 | 1 | 0% | 1,348 | 1,213 | -10% | 0 | 0 | — |
case-20 | pass→fail | 23,212 | 16,958 | -27% | 1 | 1 | 0% | 3,945 | 625 | -84% | 0 | 0 | — |
case-21 | pass→pass | 16,975 | 15,632 | -8% | 1 | 1 | 0% | 2,230 | 2,095 | -6% | 0 | 0 | — |
case-22 | fail→fail | 8,155 | 32,793 | +302% | 1 | 1 | 0% | 245 | 540 | +120% | 0 | 0 | — |
case-23 | fail→pass | 18,160 | 9,128 | -50% | 1 | 1 | 0% | 1,928 | 863 | -55% | 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. 23 cases were attempted, and 18 counted toward the lift figure. The other 5 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 +39 percentage points is the difference between those two pass rates over the 18 comparable cases. 7 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/24/2026 | -5% |
| gemini-3.6-flash | verified | 8/9/2026 | — |
Other measured skills in the registry, with their headline benchmark lift.