Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Review code changes for correctness defects, regressions, security risks, data-integrity problems, concurrency hazards, and missing meaningful tests. Use for pull requests, branches, commits, patches, or working-tree reviews; do not use for implementation unless the user separately requests fixes.
.claude/skills/hashgraph-online-review-code/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 90% | 10 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -22% | 0% |
| case-17 | ✓→✗ | ▼ Worse | -19% | 0% |
persistence, side effects, and tests.
narrow check that materially confirms a suspected issue.
Use correctness as the default lens. For large, risky, or requirement-heavy changes, add the standards and specification lenses in references/review-lenses.md. Keep each finding tied to its source; documented rules and acceptance criteria outrank personal preference.
Each finding must include:
Do not report style preferences, theoretical risks without a reachable path, or issues outside the change unless the diff materially exposes them.
without explicit authorization and verified isolation.
Read references/review-checklist.md when the change touches APIs, data, authorization, concurrency, infrastructure, or tests.
If there are no actionable findings, say so directly and name residual unverified areas. Keep the summary shorter than the findings.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | pass→fail | 4,486 | 7,230 | +61% | 1 | 1 | 0% | 786 | 617 | -22% | 0 | 0 | — |
case-01 | fail→fail | 11,518 | 13,936 | +21% | 1 | 1 | 0% | 1,578 | 631 | -60% | 0 | 0 | — |
case-02 | pass→pass | 15,466 | 10,878 | -30% | 1 | 1 | 0% | 1,591 | 1,356 | -15% | 0 | 0 | — |
case-03 | pass→pass | 12,643 | 10,183 | -19% | 1 | 1 | 0% | 1,444 | 1,207 | -16% | 0 | 0 | — |
case-04 | pass→pass | 14,701 | 5,250 | -64% | 1 | 1 | 0% | 1,479 | 1,123 | -24% | 0 | 0 | — |
case-05 | pass→pass | 15,990 | 10,302 | -36% | 1 | 1 | 0% | 1,713 | 1,217 | -29% | 0 | 0 | — |
case-06 | pass→pass | 11,079 | 15,164 | +37% | 1 | 1 | 0% | 1,667 | 1,894 | +14% | 0 | 0 | — |
case-07 | pass→pass | 20,410 | 9,919 | -51% | 1 | 1 | 0% | 2,243 | 1,289 | -43% | 0 | 0 | — |
case-08 | fail→pass | 12,690 | 8,022 | -37% | 1 | 1 | 0% | 1,179 | 827 | -30% | 0 | 0 | — |
case-09 | pass→pass | 10,959 | 10,759 | -2% | 1 | 1 | 0% | 1,757 | 1,154 | -34% | 0 | 0 | — |
case-10 | pass→pass | 18,586 | 13,532 | -27% | 1 | 1 | 0% | 1,689 | 1,727 | +2% | 0 | 0 | — |
case-11 | pass→pass | 11,599 | 10,707 | -8% | 1 | 1 | 0% | 1,482 | 1,121 | -24% | 0 | 0 | — |
case-12 | pass→pass | 14,117 | 8,073 | -43% | 1 | 1 | 0% | 1,311 | 788 | -40% | 0 | 0 | — |
case-13 | pass→pass | 14,858 | 7,464 | -50% | 1 | 1 | 0% | 2,276 | 1,406 | -38% | 0 | 0 | — |
case-14 | pass→pass | 9,749 | 8,630 | -11% | 1 | 1 | 0% | 1,465 | 933 | -36% | 0 | 0 | — |
case-15 | fail→pass | 12,247 | 10,745 | -12% | 1 | 1 | 0% | 1,244 | 1,079 | -13% | 0 | 0 | — |
case-16 | pass→pass | 10,639 | 6,895 | -35% | 1 | 1 | 0% | 1,807 | 1,560 | -14% | 0 | 0 | — |
case-17 | pass→fail | 14,093 | 15,177 | +8% | 1 | 1 | 0% | 2,177 | 1,763 | -19% | 0 | 0 | — |
case-18 | pass→pass | 14,946 | 6,215 | -58% | 1 | 1 | 0% | 1,590 | 1,389 | -13% | 0 | 0 | — |
case-19 | fail→pass | 18,960 | 11,101 | -41% | 1 | 1 | 0% | 1,851 | 1,401 | -24% | 0 | 0 | — |
case-20 | pass→pass | 12,944 | 6,504 | -50% | 1 | 1 | 0% | 1,744 | 1,455 | -17% | 0 | 0 | — |
case-22 | pass→fail | 25,821 | 9,193 | -64% | 1 | 1 | 0% | 3,987 | 561 | -86% | 0 | 0 | — |
case-23 | pass→fail | 9,479 | 9,507 | +0% | 1 | 1 | 0% | 1,488 | 594 | -60% | 0 | 0 | — |
case-24 | pass→pass | 15,302 | 27,578 | +80% | 1 | 1 | 0% | 1,941 | 5,957 | +207% | 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. 24 cases were attempted, and 20 counted toward the lift figure. The other 4 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 -4 percentage points is the difference between those two pass rates over the 20 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.