Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Web search using Jina Search API. Returns search results with titles, URLs, and descriptions. Use for finding documentation, facts, current information, or any web content. Lightweight, no browser required.
.claude/skills/dicklesworthstone-web-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -15% | 0% |
Perform web searches using the Jina Search API. Returns formatted search results with titles, URLs, and descriptions.
Optionally get a Jina API key for higher rate limits:
~/.profile or ~/.zprofile for zsh):bash export JINA_API_KEY="your-api-key-here"
Without an API key, the service works with rate limits.
bash{baseDir}/search.py "your search query"
bash# Basic search {baseDir}/search.py "python async await tutorial" # Search for recent news {baseDir}/search.py "latest AI developments 2024" # Find documentation {baseDir}/search.py "nodejs fs promises API"
Returns markdown-formatted search results:
## Search Results
[Title of first result](https://example.com/page1)
Description or snippet from the search result...
[Title of second result](https://example.com/page2)
Description or snippet from the search result...| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,083 | 13,469 | +165% | 1 | 1 | 0% | 172 | 1,066 | +520% | 0 | 0 | — |
case-02 | fail→fail | 4,477 | 5,929 | +32% | 1 | 1 | 0% | 192 | 585 | +205% | 0 | 0 | — |
case-03 | fail→fail | 9,035 | 5,521 | -39% | 1 | 1 | 0% | 1,617 | 592 | -63% | 0 | 0 | — |
case-04 | pass→pass | 4,396 | 2,195 | -50% | 1 | 1 | 0% | 784 | 668 | -15% | 0 | 0 | — |
case-05 | fail→pass | 8,030 | 3,987 | -50% | 1 | 1 | 0% | 1,473 | 975 | -34% | 0 | 0 | — |
case-06 | fail→pass | 6,575 | 1,874 | -71% | 1 | 1 | 0% | 1,028 | 649 | -37% | 0 | 0 | — |
case-07 | fail→pass | 5,903 | 2,014 | -66% | 1 | 1 | 0% | 977 | 622 | -36% | 0 | 0 | — |
case-08 | fail→fail | 8,200 | 1,808 | -78% | 1 | 1 | 0% | 1,262 | 616 | -51% | 0 | 0 | — |
case-09 | pass→pass | 10,503 | 2,261 | -78% | 1 | 1 | 0% | 1,720 | 737 | -57% | 0 | 0 | — |
case-10 | fail→pass | 5,692 | 2,502 | -56% | 1 | 1 | 0% | 930 | 734 | -21% | 0 | 0 | — |
case-11 | pass→pass | 9,506 | 2,584 | -73% | 1 | 1 | 0% | 1,571 | 667 | -58% | 0 | 0 | — |
case-12 | fail→fail | 11,427 | 5,623 | -51% | 1 | 1 | 0% | 1,680 | 575 | -66% | 0 | 0 | — |
case-13 | fail→fail | 4,581 | 6,511 | +42% | 1 | 1 | 0% | 189 | 643 | +240% | 0 | 0 | — |
case-14 | fail→fail | 9,286 | 5,307 | -43% | 1 | 1 | 0% | 1,661 | 669 | -60% | 0 | 0 | — |
case-15 | fail→fail | 5,639 | 5,733 | +2% | 1 | 1 | 0% | 225 | 626 | +178% | 0 | 0 | — |
case-16 | fail→fail | 9,479 | 3,088 | -67% | 1 | 1 | 0% | 1,520 | 768 | -49% | 0 | 0 | — |
case-17 | fail→fail | 4,256 | 9,448 | +122% | 1 | 1 | 0% | 167 | 941 | +463% | 0 | 0 | — |
case-18 | fail→fail | 19,856 | 5,743 | -71% | 1 | 1 | 0% | 2,864 | 521 | -82% | 0 | 0 | — |
case-19 | fail→fail | 13,221 | 8,525 | -36% | 1 | 1 | 0% | 1,921 | 872 | -55% | 0 | 0 | — |
case-20 | pass→pass | 8,334 | 10,300 | +24% | 1 | 1 | 0% | 1,423 | 2,083 | +46% | 0 | 0 | — |
case-21 | pass→pass | 9,767 | 24,851 | +154% | 1 | 1 | 0% | 1,795 | 2,615 | +46% | 0 | 0 | — |
case-22 | pass→pass | 11,573 | 13,200 | +14% | 1 | 1 | 0% | 468 | 1,201 | +157% | 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 12 counted toward the lift figure. The other 10 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 +18 percentage points is the difference between those two pass rates over the 12 comparable cases. 3 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.