Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when auditing slow page loads, heavy assets, or rendering delays related to Optimize third-party script loading. Verify the actual bottleneck in DevTools, Lighthouse, or field data before recommending changes.
.claude/skills/thedaviddias-third-party-scripts/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-02 | ✓→✗ | ▼ Worse | 0% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | -5% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 20% | 0% |
Third-party scripts are the #1 cause of slow websites—unoptimized loading can add 2-5 seconds to your page load time.
Analyze third-party scripts on this page to ensure they use async/defer and don't block rendering.
Add async or defer attributes to third-party scripts and consider lazy loading non-critical scripts.
Explain the performance impact of third-party scripts and strategies to minimize their blocking effect.
Review the routes, assets, and loading behavior that affect Optimize third-party script loading. Flag exact files, requests, or rendering steps that add unnecessary network, CPU, or layout cost, and describe the measurement method used to confirm the issue.
For full implementation details, code examples, and framework-specific guidance, see references/rule.md.
Rule page: https://frontendchecklist.io/en/rules/performance/third-party-scripts
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | fail→pass | 13,536 | 12,511 | -8% | 1 | 1 | 0% | 2,245 | 2,424 | +8% | 0 | 0 | — |
case-01 | pass→pass | 14,503 | 13,106 | -10% | 1 | 1 | 0% | 2,971 | 2,812 | -5% | 0 | 0 | — |
case-02 | pass→fail | 14,047 | 11,952 | -15% | 1 | 1 | 0% | 2,621 | 2,625 | +0% | 0 | 0 | — |
case-03 | pass→pass | 11,527 | 11,723 | +2% | 1 | 1 | 0% | 2,038 | 2,436 | +20% | 0 | 0 | — |
case-04 | pass→pass | 5,610 | 4,202 | -25% | 1 | 1 | 0% | 1,102 | 1,047 | -5% | 0 | 0 | — |
case-05 | pass→pass | 7,357 | 5,763 | -22% | 1 | 1 | 0% | 1,346 | 1,275 | -5% | 0 | 0 | — |
case-06 | pass→pass | 12,651 | 7,693 | -39% | 1 | 1 | 0% | 2,247 | 1,684 | -25% | 0 | 0 | — |
case-07 | pass→pass | 8,836 | 6,413 | -27% | 1 | 1 | 0% | 1,511 | 1,398 | -7% | 0 | 0 | — |
case-08 | fail→fail | 9,205 | 7,142 | -22% | 1 | 1 | 0% | 1,770 | 1,498 | -15% | 0 | 0 | — |
case-09 | pass→pass | 11,273 | 11,196 | -1% | 1 | 1 | 0% | 1,943 | 2,248 | +16% | 0 | 0 | — |
case-10 | pass→pass | 8,325 | 8,997 | +8% | 1 | 1 | 0% | 1,573 | 2,066 | +31% | 0 | 0 | — |
case-11 | pass→pass | 4,495 | 4,164 | -7% | 1 | 1 | 0% | 825 | 1,030 | +25% | 0 | 0 | — |
case-12 | pass→pass | 4,012 | 3,558 | -11% | 1 | 1 | 0% | 741 | 868 | +17% | 0 | 0 | — |
case-14 | pass→pass | 13,129 | 10,252 | -22% | 1 | 1 | 0% | 2,318 | 2,124 | -8% | 0 | 0 | — |
case-15 | fail→pass | 11,060 | 9,469 | -14% | 1 | 1 | 0% | 2,007 | 2,048 | +2% | 0 | 0 | — |
case-16 | fail→fail | 18,161 | 14,078 | -22% | 1 | 1 | 0% | 3,593 | 2,972 | -17% | 0 | 0 | — |
case-17 | pass→pass | 13,694 | 9,506 | -31% | 1 | 1 | 0% | 2,403 | 1,788 | -26% | 0 | 0 | — |
case-18 | pass→pass | 7,891 | 4,822 | -39% | 1 | 1 | 0% | 1,329 | 1,040 | -22% | 0 | 0 | — |
case-19 | pass→pass | 6,230 | 6,198 | -1% | 1 | 1 | 0% | 1,190 | 1,328 | +12% | 0 | 0 | — |
case-20 | fail→fail | 16,422 | 14,099 | -14% | 1 | 1 | 0% | 2,956 | 2,853 | -3% | 0 | 0 | — |
case-21 | pass→pass | 8,125 | 7,888 | -3% | 1 | 1 | 0% | 1,428 | 1,817 | +27% | 0 | 0 | — |
case-22 | pass→pass | 13,211 | 11,790 | -11% | 1 | 1 | 0% | 2,302 | 2,352 | +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. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.