Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Web search via self-hosted SearxNG. Aggregates Google, Bing, DuckDuckGo, Brave. Returns title/url/snippet. Zero API key required.
.claude/skills/ythx-101-ask-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 124% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 15% | 0% |
Web search powered by SearxNG. Aggregates multiple search engines, zero API key, full privacy.
bashask-search "your query" # top 10 results ask-search "query" --num 5 # limit results ask-search "AI news" --categories news # news only ask-search "query" --lang zh-CN # Chinese results ask-search "query" --urls-only # URL list (pipe to web_fetch) ask-search "query" --json # raw JSON
ask-search "topic" to get candidatesweb_fetch on the URL for full content| Flag | Short | Description | |------|-------|-------------| | --num N | -n | Max results (default 10) | | --engines | -e | google,bing,duckduckgo,brave | | --lang | -l | zh-CN, en, ja, ko | | --categories | -c | general,news,images,science | | --json | -j | Raw JSON output | | --urls-only | -u | URLs only |
Requires SearxNG running locally. Set SEARXNG_URL if not on default port 8080.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→fail | 13,096 | 14,745 | +13% | 1 | 1 | 0% | 2,537 | 628 | -75% | 0 | 0 | — |
case-01 | fail→fail | 10,505 | 10,435 | -1% | 1 | 1 | 0% | 1,697 | 593 | -65% | 0 | 0 | — |
case-02 | fail→fail | 3,290 | 3,432 | +4% | 1 | 1 | 0% | 685 | 539 | -21% | 0 | 0 | — |
case-03 | fail→pass | 3,926 | 1,384 | -65% | 1 | 1 | 0% | 249 | 559 | +124% | 0 | 0 | — |
case-04 | fail→pass | 9,656 | 2,056 | -79% | 1 | 1 | 0% | 1,809 | 747 | -59% | 0 | 0 | — |
case-05 | pass→pass | 5,404 | 1,542 | -71% | 1 | 1 | 0% | 951 | 603 | -37% | 0 | 0 | — |
case-07 | fail→pass | 40,740 | 2,015 | -95% | 1 | 1 | 0% | 1,436 | 543 | -62% | 0 | 0 | — |
case-08 | fail→fail | 10,938 | 5,576 | -49% | 1 | 1 | 0% | 2,398 | 602 | -75% | 0 | 0 | — |
case-09 | fail→pass | 7,046 | 1,802 | -74% | 1 | 1 | 0% | 1,277 | 608 | -52% | 0 | 0 | — |
case-10 | fail→pass | 6,497 | 5,246 | -19% | 1 | 1 | 0% | 679 | 779 | +15% | 0 | 0 | — |
case-11 | fail→pass | 9,981 | 1,312 | -87% | 1 | 1 | 0% | 1,750 | 536 | -69% | 0 | 0 | — |
case-12 | fail→pass | 10,512 | 1,316 | -87% | 1 | 1 | 0% | 1,873 | 541 | -71% | 0 | 0 | — |
case-13 | pass→pass | 5,614 | 1,874 | -67% | 1 | 1 | 0% | 931 | 692 | -26% | 0 | 0 | — |
case-14 | pass→pass | 5,962 | 1,473 | -75% | 1 | 1 | 0% | 1,184 | 580 | -51% | 0 | 0 | — |
case-15 | fail→pass | 4,403 | 5,060 | +15% | 1 | 1 | 0% | 211 | 769 | +264% | 0 | 0 | — |
case-16 | fail→pass | 11,729 | 3,366 | -71% | 1 | 1 | 0% | 2,030 | 688 | -66% | 0 | 0 | — |
case-17 | fail→fail | 14,849 | 4,711 | -68% | 1 | 1 | 0% | 2,222 | 531 | -76% | 0 | 0 | — |
case-18 | fail→pass | 4,292 | 1,172 | -73% | 1 | 1 | 0% | 743 | 499 | -33% | 0 | 0 | — |
case-19 | pass→pass | 7,913 | 1,716 | -78% | 1 | 1 | 0% | 1,221 | 574 | -53% | 0 | 0 | — |
case-20 | pass→pass | 3,659 | 1,896 | -48% | 1 | 1 | 0% | 572 | 635 | +11% | 0 | 0 | — |
case-21 | pass→pass | 9,390 | 7,780 | -17% | 1 | 1 | 0% | 1,762 | 1,835 | +4% | 0 | 0 | — |
case-22 | pass→pass | 7,856 | 4,424 | -44% | 1 | 1 | 0% | 1,529 | 1,172 | -23% | 0 | 0 | — |
case-23 | pass→pass | 14,591 | 10,938 | -25% | 1 | 1 | 0% | 2,906 | 2,577 | -11% | 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. 23 cases were attempted, and 16 counted toward the lift figure. The other 7 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 +43 percentage points is the difference between those two pass rates over the 16 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.