Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audit and optimize an existing Convex app: security, scale, upgrades, observability.
.claude/skills/get-convex-convex-optimize/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -28% | 0% |
<!-- GENERATED from convex-agents content/capabilities/optimize.json — do not edit by hand. -->
The remediation WORKFLOW for an existing app: open with a scored assessment, then act on it — upgrade stale components and set up observability — plan-then-confirm-then-apply. The assessment itself is delegated to launch-readiness (the findings-bus scorer); optimize's distinct value is the actions it takes on the result.
convex/ directory, the schema, and whether it's an anonymous or cloud deployment.launch-readiness — one scored, deduped report across authz/reviewer/advisor/insights with an ordered fix plan. Do not re-run those passes by hand; optimize consumes launch-readiness's report rather than re-implementing the audit.check-updates against the pinned @convex-dev/* components and fold stale-component (staleness-class) findings into the same plan.sentinel.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 5,166 | 6,356 | +23% | 1 | 1 | 0% | 698 | 1,408 | +102% | 0 | 0 | — |
case-01 | fail→fail | 9,110 | 6,882 | -24% | 1 | 1 | 0% | 499 | 855 | +71% | 0 | 0 | — |
case-02 | fail→fail | 3,580 | 3,326 | -7% | 1 | 1 | 0% | 307 | 778 | +153% | 0 | 0 | — |
case-03 | fail→fail | 5,186 | 4,706 | -9% | 1 | 1 | 0% | 243 | 954 | +293% | 0 | 0 | — |
case-04 | pass→pass | 13,416 | 6,915 | -48% | 1 | 1 | 0% | 1,322 | 1,395 | +6% | 0 | 0 | — |
case-05 | pass→pass | 17,211 | 8,367 | -51% | 1 | 1 | 0% | 2,339 | 1,699 | -27% | 0 | 0 | — |
case-06 | fail→pass | 10,030 | 4,096 | -59% | 1 | 1 | 0% | 1,332 | 1,070 | -20% | 0 | 0 | — |
case-07 | pass→pass | 4,998 | 4,116 | -18% | 1 | 1 | 0% | 784 | 979 | +25% | 0 | 0 | — |
case-08 | fail→pass | 6,226 | 2,260 | -64% | 1 | 1 | 0% | 935 | 766 | -18% | 0 | 0 | — |
case-10 | pass→pass | 12,951 | 3,930 | -70% | 1 | 1 | 0% | 2,217 | 1,014 | -54% | 0 | 0 | — |
case-11 | fail→pass | 13,285 | 4,553 | -66% | 1 | 1 | 0% | 1,992 | 1,213 | -39% | 0 | 0 | — |
case-12 | fail→pass | 14,599 | 5,131 | -65% | 1 | 1 | 0% | 2,083 | 1,059 | -49% | 0 | 0 | — |
case-13 | fail→pass | 10,028 | 5,262 | -48% | 1 | 1 | 0% | 1,500 | 1,076 | -28% | 0 | 0 | — |
case-14 | fail→pass | 15,716 | 5,603 | -64% | 1 | 1 | 0% | 2,053 | 1,191 | -42% | 0 | 0 | — |
case-15 | fail→fail | 8,730 | 2,846 | -67% | 1 | 1 | 0% | 1,150 | 809 | -30% | 0 | 0 | — |
case-16 | fail→pass | 15,201 | 8,981 | -41% | 1 | 1 | 0% | 2,403 | 1,533 | -36% | 0 | 0 | — |
case-17 | pass→pass | 6,323 | 5,030 | -20% | 1 | 1 | 0% | 1,121 | 1,089 | -3% | 0 | 0 | — |
case-18 | pass→pass | 8,376 | 3,207 | -62% | 1 | 1 | 0% | 1,262 | 833 | -34% | 0 | 0 | — |
case-19 | pass→pass | 9,570 | 4,781 | -50% | 1 | 1 | 0% | 1,200 | 1,226 | +2% | 0 | 0 | — |
case-20 | pass→pass | 15,479 | 15,010 | -3% | 1 | 1 | 0% | 2,930 | 3,318 | +13% | 0 | 0 | — |
case-21 | pass→pass | 18,828 | 15,057 | -20% | 1 | 1 | 0% | 2,817 | 2,746 | -3% | 0 | 0 | — |
case-22 | pass→pass | 9,277 | 8,011 | -14% | 1 | 1 | 0% | 1,755 | 1,904 | +8% | 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 20 counted toward the lift figure. The other 2 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 +32 percentage points is the difference between those two pass rates over the 20 comparable cases.
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.