Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Standard workflow to address pull request review comments in Cecil. Use when responding to review feedback, fixing requested changes, or preparing a clear reviewer reply with reproduction, fix, tests, and response.
.claude/skills/cecilapp-address-pr-comments-cecil/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -37% | 0% |
Use this workflow to process review feedback consistently.
Response profile: Standard.
AGENTS.md and .github/instructions/php-src.instructions.md.composer code:style, composer code:analyse, composer test, composer test:cli.Use this structure in your final PR comment response:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 4,397 | 6,620 | +51% | 1 | 1 | 0% | 234 | 530 | +126% | 0 | 0 | — |
case-02 | fail→fail | 3,767 | 4,193 | +11% | 1 | 1 | 0% | 128 | 600 | +369% | 0 | 0 | — |
case-03 | fail→fail | 5,695 | 4,186 | -26% | 1 | 1 | 0% | 299 | 660 | +121% | 0 | 0 | — |
case-04 | pass→pass | 15,939 | 13,851 | -13% | 1 | 1 | 0% | 2,527 | 2,190 | -13% | 0 | 0 | — |
case-05 | pass→fail | 7,582 | 7,045 | -7% | 1 | 1 | 0% | 1,232 | 781 | -37% | 0 | 0 | — |
case-06 | fail→fail | 7,796 | 5,304 | -32% | 1 | 1 | 0% | 1,077 | 632 | -41% | 0 | 0 | — |
case-07 | pass→pass | 6,846 | 3,653 | -47% | 1 | 1 | 0% | 1,281 | 1,013 | -21% | 0 | 0 | — |
case-13 | fail→pass | 6,954 | 1,921 | -72% | 1 | 1 | 0% | 966 | 667 | -31% | 0 | 0 | — |
case-08 | fail→fail | 9,599 | 2,322 | -76% | 1 | 1 | 0% | 1,749 | 809 | -54% | 0 | 0 | — |
case-09 | pass→pass | 7,817 | 3,427 | -56% | 1 | 1 | 0% | 1,301 | 963 | -26% | 0 | 0 | — |
case-10 | pass→pass | 8,822 | 1,824 | -79% | 1 | 1 | 0% | 1,535 | 676 | -56% | 0 | 0 | — |
case-11 | pass→pass | 10,085 | 5,875 | -42% | 1 | 1 | 0% | 1,635 | 1,309 | -20% | 0 | 0 | — |
case-12 | fail→pass | 10,702 | 3,105 | -71% | 1 | 1 | 0% | 1,616 | 936 | -42% | 0 | 0 | — |
case-14 | fail→fail | 10,237 | 1,818 | -82% | 1 | 1 | 0% | 1,431 | 644 | -55% | 0 | 0 | — |
case-15 | fail→pass | 9,781 | 4,644 | -53% | 1 | 1 | 0% | 1,524 | 1,095 | -28% | 0 | 0 | — |
case-16 | pass→pass | 10,621 | 2,794 | -74% | 1 | 1 | 0% | 1,592 | 779 | -51% | 0 | 0 | — |
case-17 | pass→pass | 9,454 | 3,269 | -65% | 1 | 1 | 0% | 1,380 | 910 | -34% | 0 | 0 | — |
case-18 | pass→pass | 8,317 | 3,146 | -62% | 1 | 1 | 0% | 1,246 | 871 | -30% | 0 | 0 | — |
case-19 | fail→pass | 12,042 | 5,966 | -50% | 1 | 1 | 0% | 1,978 | 1,361 | -31% | 0 | 0 | — |
case-20 | pass→pass | 6,309 | 2,172 | -66% | 1 | 1 | 0% | 832 | 663 | -20% | 0 | 0 | — |
case-21 | pass→pass | 10,721 | 5,495 | -49% | 1 | 1 | 0% | 1,582 | 1,177 | -26% | 0 | 0 | — |
case-22 | pass→pass | 6,299 | 4,160 | -34% | 1 | 1 | 0% | 984 | 1,006 | +2% | 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, and 18 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 +14 percentage points is the difference between those two pass rates over the 18 comparable cases. 2 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.