Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Optimize CodeRabbit review speed, relevance, and signal-to-noise ratio. Use when reviews take too long, contain too many irrelevant comments, or when teams are experiencing review fatigue. Trigger with phrases like "coderabbit performance", "optimize coderabbit", "coderabbit slow", "coderabbit noise", "coderabbit too many comments", "coderabbit relevance".
.claude/skills/jeremylongshore-coderabbit-performance-tuning/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 123% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 244% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 80% | 0% |
Optimize operator-controlled inputs and measure effects. Separate service latency, provider delay, limits, diff size, paths, tools, and guidance.
references/official-docs.md and re-check any time-sensitive contract before execution.Treat Git-provider sessions, CodeRabbit web sessions, CLI credentials, and CodeRabbit API keys as separate credentials. Use only an already-approved session or secret-manager reference, never print a secret, and do not place credentials in .coderabbit.yaml, source files, logs, or deliverables.
Require security or code-owner approval before exclusions, tool disablement, or shared changes. Keep analysis and drafts local until approval is explicit, and record who approved the action and its scope.
A baseline, hypothesis, patch, comparison, risk check, and keep-or-rollback decision. Include source dates, unknowns, and the exact boundary between observed fact and recommendation.
| Condition | Response | |---|---| | Current contract is unclear or docs disagree | Stop mutation, cite both sources, and request owner resolution. | | Required access or approval is missing | Produce a draft and evidence plan only. | | Validation or pilot behavior differs from expectation | Restore the prior state and retain the failed evidence. | | Output contains secrets or private code | Stop, quarantine the artifact, redact it, and notify the data owner. |
Exclude generated artifacts while retaining sensitive code.
Tune incremental-review pause for a high-churn branch.
references/official-docs.md.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 4,841 | 2,597 | -46% | 1 | 1 | 0% | 881 | 2,652 | +201% | 0 | 0 | — |
case-11 | fail→pass | 11,709 | 5,079 | -57% | 1 | 1 | 0% | 2,084 | 3,070 | +47% | 0 | 0 | — |
case-12 | pass→pass | 10,850 | 1,970 | -82% | 1 | 1 | 0% | 1,616 | 2,403 | +49% | 0 | 0 | — |
case-05 | pass→pass | 4,297 | 2,497 | -42% | 1 | 1 | 0% | 846 | 2,559 | +202% | 0 | 0 | — |
case-01 | fail→fail | 11,646 | 8,341 | -28% | 1 | 1 | 0% | 2,216 | 3,776 | +70% | 0 | 0 | — |
case-02 | fail→pass | 16,084 | 9,806 | -39% | 1 | 1 | 0% | 3,447 | 4,216 | +22% | 0 | 0 | — |
case-03 | pass→pass | 5,574 | 3,043 | -45% | 1 | 1 | 0% | 922 | 2,579 | +180% | 0 | 0 | — |
case-04 | pass→pass | 8,264 | 3,709 | -55% | 1 | 1 | 0% | 1,231 | 2,762 | +124% | 0 | 0 | — |
case-07 | pass→pass | 4,614 | 3,028 | -34% | 1 | 1 | 0% | 913 | 2,793 | +206% | 0 | 0 | — |
case-08 | fail→pass | 7,503 | 3,819 | -49% | 1 | 1 | 0% | 1,311 | 2,921 | +123% | 0 | 0 | — |
case-09 | fail→pass | 4,707 | 3,024 | -36% | 1 | 1 | 0% | 794 | 2,733 | +244% | 0 | 0 | — |
case-10 | pass→pass | 3,046 | 2,413 | -21% | 1 | 1 | 0% | 568 | 2,617 | +361% | 0 | 0 | — |
case-13 | fail→pass | 9,091 | 3,260 | -64% | 1 | 1 | 0% | 1,527 | 2,745 | +80% | 0 | 0 | — |
case-14 | pass→pass | 5,743 | 3,177 | -45% | 1 | 1 | 0% | 1,018 | 2,701 | +165% | 0 | 0 | — |
case-15 | pass→pass | 7,005 | 5,013 | -28% | 1 | 1 | 0% | 1,255 | 2,723 | +117% | 0 | 0 | — |
case-16 | pass→pass | 4,706 | 1,893 | -60% | 1 | 1 | 0% | 747 | 2,449 | +228% | 0 | 0 | — |
case-17 | fail→pass | 15,577 | 7,957 | -49% | 1 | 1 | 0% | 2,657 | 3,551 | +34% | 0 | 0 | — |
case-18 | fail→pass | 9,809 | 3,010 | -69% | 1 | 1 | 0% | 1,716 | 2,653 | +55% | 0 | 0 | — |
case-19 | pass→pass | 5,335 | 2,219 | -58% | 1 | 1 | 0% | 1,026 | 2,560 | +150% | 0 | 0 | — |
case-20 | pass→pass | 10,573 | 7,975 | -25% | 1 | 1 | 0% | 2,038 | 3,768 | +85% | 0 | 0 | — |
case-21 | pass→pass | 6,268 | 6,346 | +1% | 1 | 1 | 0% | 1,291 | 3,390 | +163% | 0 | 0 | — |
case-22 | pass→pass | 9,332 | 5,482 | -41% | 1 | 1 | 0% | 1,802 | 3,193 | +77% | 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 +32 percentage points is the difference between those two pass rates over the 22 comparable cases.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.