Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Import SOP: deep full-page content analysis (from web-browsing skill)
.claude/skills/yogsoth-ai-experiment-execution-web-research/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-01 | ✓→✗ | ▼ Worse | -85% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -65% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -63% | 0% |
| case-21 | ✓→✓ | ＝ Same ✓ | 29% | 0% |
Deep full-page content analysis — imported from web-browsing skill.
Import — invokes web-browsing skill's web-research SOP.
Available to all campaigns for in-depth reading and analysis of specific web pages (technical docs, framework guides, API documentation, etc.).
rag-web-browserbrave_llm_context<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | web-research | Deep web research — fetches full page content for analysis. Snippets alone are PROHIBITED for conclusions. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→fail | 21,470 | 34,751 | +62% | 1 | 1 | 0% | 3,783 | 556 | -85% | 0 | 0 | — |
case-02 | fail→fail | 15,001 | 6,102 | -59% | 1 | 1 | 0% | 2,566 | 517 | -80% | 0 | 0 | — |
case-03 | fail→fail | 21,296 | 7,018 | -67% | 1 | 1 | 0% | 3,448 | 585 | -83% | 0 | 0 | — |
case-04 | fail→fail | 24,972 | 10,080 | -60% | 1 | 1 | 0% | 3,818 | 806 | -79% | 0 | 0 | — |
case-05 | fail→fail | 14,335 | 8,539 | -40% | 1 | 1 | 0% | 2,235 | 693 | -69% | 0 | 0 | — |
case-06 | pass→fail | 13,396 | 8,125 | -39% | 1 | 1 | 0% | 1,666 | 584 | -65% | 0 | 0 | — |
case-07 | fail→fail | 22,026 | 8,911 | -60% | 1 | 1 | 0% | 3,348 | 836 | -75% | 0 | 0 | — |
case-08 | fail→fail | 16,263 | 7,461 | -54% | 1 | 1 | 0% | 2,246 | 562 | -75% | 0 | 0 | — |
case-09 | fail→fail | 5,615 | 6,415 | +14% | 1 | 1 | 0% | 309 | 604 | +95% | 0 | 0 | — |
case-10 | fail→fail | 21,279 | 8,231 | -61% | 1 | 1 | 0% | 3,572 | 565 | -84% | 0 | 0 | — |
case-11 | fail→fail | 6,679 | 7,898 | +18% | 1 | 1 | 0% | 274 | 578 | +111% | 0 | 0 | — |
case-12 | fail→pass | 13,369 | 8,442 | -37% | 1 | 1 | 0% | 2,685 | 1,696 | -37% | 0 | 0 | — |
case-13 | fail→fail | 22,853 | 7,964 | -65% | 1 | 1 | 0% | 3,329 | 699 | -79% | 0 | 0 | — |
case-14 | fail→fail | 5,001 | 6,281 | +26% | 1 | 1 | 0% | 242 | 576 | +138% | 0 | 0 | — |
case-15 | pass→fail | 11,117 | 7,123 | -36% | 1 | 1 | 0% | 1,649 | 605 | -63% | 0 | 0 | — |
case-16 | fail→fail | 15,191 | 8,326 | -45% | 1 | 1 | 0% | 2,313 | 1,513 | -35% | 0 | 0 | — |
case-17 | fail→fail | 7,721 | 6,701 | -13% | 1 | 1 | 0% | 383 | 548 | +43% | 0 | 0 | — |
case-18 | fail→fail | 18,354 | 6,721 | -63% | 1 | 1 | 0% | 3,075 | 513 | -83% | 0 | 0 | — |
case-19 | fail→fail | 16,673 | 8,823 | -47% | 1 | 1 | 0% | 2,527 | 686 | -73% | 0 | 0 | — |
case-20 | fail→fail | 8,944 | 3,280 | -63% | 1 | 1 | 0% | 1,462 | 686 | -53% | 0 | 0 | — |
case-21 | pass→pass | 12,281 | 13,856 | +13% | 1 | 1 | 0% | 2,215 | 2,860 | +29% | 0 | 0 | — |
case-22 | fail→fail | 4,987 | 3,283 | -34% | 1 | 1 | 0% | 721 | 647 | -10% | 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 5 counted toward the lift figure. The other 17 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 -9 percentage points is the difference between those two pass rates over the 5 comparable cases. 8 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.