Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Review only changes since last commit using impact analysis. Token-efficient delta review with automatic blast-radius detection.
.claude/skills/tirth8205-review-delta/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -38% | 0% |
Review only the changed code and its blast radius.
get_minimal_context_tool(task="review changes"). If it returns status: not_ready, call build_or_update_graph_tool() and continue.detect_changes_tool(detail_level="minimal") for risk-scored changed functions, test gaps and affected flows. Changes come from git diff against HEAD~1; if the argument names a file, pass it in changed_files.get_review_context_tool(detail_level="minimal") when you need source snippets for the changed areas and review_guidance (untested functions, wide blast radius, inheritance changes).query_graph_tool(pattern="tests_for", target="<function>").detail_level="standard" only for a high-risk item the minimal output leaves unclear. Do not load whole files unless a snippet is not enough.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,511 | 11,395 | +225% | 1 | 1 | 0% | 464 | 493 | +6% | 0 | 0 | — |
case-02 | fail→fail | 8,943 | 15,276 | +71% | 1 | 1 | 0% | 1,301 | 455 | -65% | 0 | 0 | — |
case-03 | fail→fail | 9,109 | 16,438 | +80% | 1 | 1 | 0% | 609 | 437 | -28% | 0 | 0 | — |
case-04 | fail→fail | 18,366 | 41,791 | +128% | 1 | 1 | 0% | 1,930 | 630 | -67% | 0 | 0 | — |
case-05 | fail→fail | 17,949 | 12,969 | -28% | 1 | 1 | 0% | 1,823 | 652 | -64% | 0 | 0 | — |
case-06 | fail→fail | 45,307 | 3,064 | -93% | 1 | 1 | 0% | 1,181 | 641 | -46% | 0 | 0 | — |
case-07 | fail→pass | 14,878 | 13,536 | -9% | 1 | 1 | 0% | 1,625 | 704 | -57% | 0 | 0 | — |
case-08 | pass→fail | 18,252 | 12,725 | -30% | 1 | 1 | 0% | 1,879 | 732 | -61% | 0 | 0 | — |
case-09 | fail→fail | 14,968 | 8,762 | -41% | 1 | 1 | 0% | 1,309 | 682 | -48% | 0 | 0 | — |
case-10 | pass→pass | 15,240 | 15,421 | +1% | 1 | 1 | 0% | 2,216 | 1,159 | -48% | 0 | 0 | — |
case-11 | fail→fail | 18,234 | 36,312 | +99% | 1 | 1 | 0% | 1,977 | 624 | -68% | 0 | 0 | — |
case-12 | pass→fail | 19,406 | 7,441 | -62% | 1 | 1 | 0% | 2,321 | 591 | -75% | 0 | 0 | — |
case-13 | fail→pass | 16,328 | 3,731 | -77% | 1 | 1 | 0% | 1,575 | 630 | -60% | 0 | 0 | — |
case-14 | fail→fail | 14,308 | 29,589 | +107% | 1 | 1 | 0% | 1,267 | 644 | -49% | 0 | 0 | — |
case-15 | fail→pass | 14,905 | 8,951 | -40% | 1 | 1 | 0% | 1,208 | 645 | -47% | 0 | 0 | — |
case-16 | pass→pass | 14,370 | 7,871 | -45% | 1 | 1 | 0% | 1,624 | 578 | -64% | 0 | 0 | — |
case-17 | pass→pass | 37,379 | 18,261 | -51% | 1 | 1 | 0% | 2,247 | 837 | -63% | 0 | 0 | — |
case-18 | fail→pass | 40,640 | 12,056 | -70% | 1 | 1 | 0% | 2,385 | 723 | -70% | 0 | 0 | — |
case-19 | pass→pass | 12,495 | 8,188 | -34% | 1 | 1 | 0% | 1,165 | 547 | -53% | 0 | 0 | — |
case-20 | pass→pass | 12,173 | 5,562 | -54% | 1 | 1 | 0% | 922 | 903 | -2% | 0 | 0 | — |
case-21 | pass→pass | 23,907 | 16,494 | -31% | 1 | 1 | 0% | 2,512 | 1,224 | -51% | 0 | 0 | — |
case-22 | fail→pass | 15,177 | 10,672 | -30% | 1 | 1 | 0% | 1,528 | 954 | -38% | 0 | 0 | — |
case-23 | fail→fail | 18,585 | 39,403 | +112% | 1 | 1 | 0% | 1,235 | 1,472 | +19% | 0 | 0 | — |
case-24 | fail→fail | 13,073 | 22,163 | +70% | 1 | 1 | 0% | 2,020 | 448 | -78% | 0 | 0 | — |
case-25 | fail→fail | 8,899 | 15,356 | +73% | 1 | 1 | 0% | 468 | 465 | -1% | 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. 25 cases were attempted, and 14 counted toward the lift figure. The other 11 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 +12 percentage points is the difference between those two pass rates over the 14 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/9/2026 | +5% |
Other measured skills in the registry, with their headline benchmark lift.