Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when auditing slow first responses on SSR pages, APIs, or cacheable HTML. Distinguish origin compute time from network latency and CDN cache misses before proposing a fix.
.claude/skills/thedaviddias-ttfb/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 153% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -22% | 0% |
TTFB is the foundation of page performance; if the server is slow to respond, all subsequent loading stages are delayed.
Review the page's TTFB and identify any bottlenecks in the server-side processing or network transmission.
Optimize the server's response time by improving database queries, caching responses, and using a CDN.
Explain how TTFB impacts the entire page load process and why it is a critical starting point for performance optimization.
Review server-rendered routes, API handlers, cache configuration, and database access on the critical path. Flag uncached HTML, slow synchronous work before the first byte, repeated upstream fetches, or expensive queries that keep TTFB above roughly 800ms.
For full implementation details, code examples, and framework-specific guidance, see references/rule.md.
Rule page: https://frontendchecklist.io/en/rules/performance/ttfb
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 10,356 | 7,031 | -32% | 1 | 1 | 0% | 1,945 | 1,532 | -21% | 0 | 0 | — |
case-01 | fail→pass | 6,644 | 14,232 | +114% | 1 | 1 | 0% | 1,152 | 2,913 | +153% | 0 | 0 | — |
case-02 | fail→pass | 13,938 | 13,084 | -6% | 1 | 1 | 0% | 2,165 | 2,614 | +21% | 0 | 0 | — |
case-03 | pass→pass | 13,637 | 9,322 | -32% | 1 | 1 | 0% | 2,253 | 1,783 | -21% | 0 | 0 | — |
case-05 | pass→pass | 20,154 | 15,532 | -23% | 1 | 1 | 0% | 3,347 | 2,823 | -16% | 0 | 0 | — |
case-06 | pass→pass | 15,745 | 12,045 | -23% | 1 | 1 | 0% | 2,878 | 2,568 | -11% | 0 | 0 | — |
case-07 | pass→pass | 18,296 | 10,181 | -44% | 1 | 1 | 0% | 3,148 | 2,134 | -32% | 0 | 0 | — |
case-08 | pass→pass | 14,477 | 12,171 | -16% | 1 | 1 | 0% | 2,526 | 2,413 | -4% | 0 | 0 | — |
case-09 | pass→pass | 14,851 | 12,626 | -15% | 1 | 1 | 0% | 2,580 | 2,416 | -6% | 0 | 0 | — |
case-10 | fail→pass | 10,327 | 9,469 | -8% | 1 | 1 | 0% | 1,649 | 1,776 | +8% | 0 | 0 | — |
case-11 | fail→pass | 15,395 | 16,174 | +5% | 1 | 1 | 0% | 2,614 | 3,043 | +16% | 0 | 0 | — |
case-12 | pass→pass | 18,797 | 24,221 | +29% | 1 | 1 | 0% | 3,048 | 3,493 | +15% | 0 | 0 | — |
case-13 | pass→pass | 17,666 | 11,684 | -34% | 1 | 1 | 0% | 2,889 | 2,444 | -15% | 0 | 0 | — |
case-14 | pass→pass | 9,077 | 6,563 | -28% | 1 | 1 | 0% | 1,423 | 1,345 | -5% | 0 | 0 | — |
case-15 | pass→pass | 18,827 | 75,347 | +300% | 1 | 1 | 0% | 2,937 | 2,579 | -12% | 0 | 0 | — |
case-16 | pass→pass | 10,093 | 9,016 | -11% | 1 | 1 | 0% | 1,741 | 1,578 | -9% | 0 | 0 | — |
case-17 | pass→pass | 15,189 | 12,307 | -19% | 1 | 1 | 0% | 2,579 | 2,338 | -9% | 0 | 0 | — |
case-18 | pass→pass | 13,424 | 12,003 | -11% | 1 | 1 | 0% | 2,390 | 2,456 | +3% | 0 | 0 | — |
case-19 | pass→pass | 15,887 | 12,949 | -18% | 1 | 1 | 0% | 2,713 | 2,586 | -5% | 0 | 0 | — |
case-20 | fail→pass | 14,039 | 10,409 | -26% | 1 | 1 | 0% | 2,442 | 1,894 | -22% | 0 | 0 | — |
case-21 | pass→pass | 17,289 | 12,539 | -27% | 1 | 1 | 0% | 2,956 | 2,348 | -21% | 0 | 0 | — |
case-22 | pass→pass | 19,052 | 15,753 | -17% | 1 | 1 | 0% | 3,108 | 3,080 | -1% | 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 +23 percentage points is the difference between those two pass rates over the 22 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.