Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Full-page web reading for non-academic perspectives — blogs, tech reports, product pages, industry analysis. Import of web-browsing/web-research skill. Must fetch full page via apify for every analyzed page.
.claude/skills/yogsoth-ai-knowledge-acquisition-web-research/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -81% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 40% | 0% |
| case-10 | ✓→✓ | ＝ Same ✓ | -74% | 0% |
Full-page web reading for non-academic perspectives (blogs, tech reports, products, industry analysis).
Import — strictly follow web-browsing/web-research skill protocol.
Must fetch full page content via apify rag-web-browser for every analyzed page. No conclusions from snippets or partial content.
Quantity target is set by the calling strategy's budget table. This SOP executes one unit = one full page read and analyzed.
web-browsing repo → skills/web-research/SKILL.md
<!-- 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 | fail→fail | 39,184 | 20,296 | -48% | 1 | 1 | 0% | 5,356 | 877 | -84% | 0 | 0 | — |
case-02 | fail→fail | 51,330 | 18,716 | -64% | 1 | 1 | 0% | 7,589 | 761 | -90% | 0 | 0 | — |
case-03 | fail→fail | 35,863 | 40,768 | +14% | 1 | 1 | 0% | 551 | 788 | +43% | 0 | 0 | — |
case-04 | pass→fail | 34,073 | 97,778 | +187% | 1 | 1 | 0% | 5,487 | 7,655 | +40% | 0 | 0 | — |
case-05 | fail→fail | 84,962 | 5,522 | -94% | 1 | 1 | 0% | 1,264 | 1,111 | -12% | 0 | 0 | — |
case-06 | fail→fail | 33,374 | 37,816 | +13% | 1 | 1 | 0% | 572 | 1,614 | +182% | 0 | 0 | — |
case-07 | fail→fail | 18,150 | 27,172 | +50% | 1 | 1 | 0% | 1,939 | 980 | -49% | 0 | 0 | — |
case-08 | fail→pass | 34,382 | 10,479 | -70% | 1 | 1 | 0% | 2,733 | 897 | -67% | 0 | 0 | — |
case-09 | fail→pass | 19,344 | 37,765 | +95% | 1 | 1 | 0% | 2,011 | 920 | -54% | 0 | 0 | — |
case-10 | pass→pass | 40,065 | 8,298 | -79% | 1 | 1 | 0% | 2,104 | 557 | -74% | 0 | 0 | — |
case-11 | pass→pass | 36,658 | 14,796 | -60% | 1 | 1 | 0% | 1,729 | 1,209 | -30% | 0 | 0 | — |
case-12 | fail→fail | 40,793 | 25,528 | -37% | 1 | 1 | 0% | 1,637 | 615 | -62% | 0 | 0 | — |
case-13 | fail→fail | 98,784 | 67,387 | -32% | 1 | 1 | 0% | 426 | 552 | +30% | 0 | 0 | — |
case-14 | fail→fail | 34,855 | 44,937 | +29% | 1 | 1 | 0% | 3,662 | 776 | -79% | 0 | 0 | — |
case-15 | fail→pass | 18,919 | 14,974 | -21% | 1 | 1 | 0% | 2,493 | 480 | -81% | 0 | 0 | — |
case-16 | fail→fail | 20,555 | 40,646 | +98% | 1 | 1 | 0% | 416 | 5,875 | +1312% | 0 | 0 | — |
case-17 | fail→fail | 44,506 | 17,855 | -60% | 1 | 1 | 0% | 6,645 | 580 | -91% | 0 | 0 | — |
case-18 | fail→fail | 18,882 | 61,126 | +224% | 1 | 1 | 0% | 466 | 1,387 | +198% | 0 | 0 | — |
case-19 | fail→fail | 21,532 | 25,411 | +18% | 1 | 1 | 0% | 3,357 | 879 | -74% | 0 | 0 | — |
case-20 | fail→fail | 27,708 | 20,440 | -26% | 1 | 1 | 0% | 3,484 | 594 | -83% | 0 | 0 | — |
case-21 | fail→fail | 45,828 | 20,270 | -56% | 1 | 1 | 0% | 6,314 | 602 | -90% | 0 | 0 | — |
case-22 | fail→fail | 43,410 | 13,732 | -68% | 1 | 1 | 0% | 4,460 | 755 | -83% | 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 10 counted toward the lift figure. The other 12 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 10 comparable cases. 2 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.