Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →This skill should be used when user asks to "search the web", "fetch content from URL", "extract page content", "use Tavily search", "scrape this website", "get information from this link", or "web search for X".
.claude/skills/fcakyon-tavily-usage/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -73% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -37% | 0% |
Use Tavily MCP tools for web search and content retrieval operations.
mcp__tavily__tavily_search)Use for:
Best for: Initial research, finding sources, broad queries
mcp__tavily__tavily-extract)Use for:
Best for: In-depth analysis, specific URL content, detailed information
tavily_extract_to_advanced.py hook automatically upgrades extract calls to advanced mode for better accuracy when needed.
mcp__tavily__tavily_search for discovery phasemcp__tavily__tavily-extract for detailed content on specific URLsTavily MCP requires:
TAVILY_API_KEY - API key from Tavily (tvly-...)Configure in shell before using the plugin.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,113 | 5,684 | -7% | 1 | 1 | 0% | 342 | 581 | +70% | 0 | 0 | — |
case-02 | fail→fail | 5,378 | 5,508 | +2% | 1 | 1 | 0% | 949 | 631 | -34% | 0 | 0 | — |
case-03 | fail→fail | 6,384 | 5,077 | -20% | 1 | 1 | 0% | 378 | 632 | +67% | 0 | 0 | — |
case-04 | fail→pass | 3,279 | 2,352 | -28% | 1 | 1 | 0% | 529 | 770 | +46% | 0 | 0 | — |
case-05 | fail→pass | 7,270 | 2,479 | -66% | 1 | 1 | 0% | 1,260 | 767 | -39% | 0 | 0 | — |
case-06 | pass→pass | 2,185 | 1,200 | -45% | 1 | 1 | 0% | 348 | 532 | +53% | 0 | 0 | — |
case-07 | pass→pass | 5,125 | 3,028 | -41% | 1 | 1 | 0% | 244 | 510 | +109% | 0 | 0 | — |
case-08 | fail→pass | 28,230 | 1,273 | -95% | 1 | 1 | 0% | 1,907 | 515 | -73% | 0 | 0 | — |
case-09 | pass→pass | 6,559 | 1,977 | -70% | 1 | 1 | 0% | 1,150 | 665 | -42% | 0 | 0 | — |
case-10 | pass→pass | 9,350 | 2,637 | -72% | 1 | 1 | 0% | 1,701 | 624 | -63% | 0 | 0 | — |
case-11 | pass→pass | 16,659 | 3,598 | -78% | 1 | 1 | 0% | 2,652 | 1,065 | -60% | 0 | 0 | — |
case-12 | pass→pass | 7,795 | 3,212 | -59% | 1 | 1 | 0% | 1,364 | 880 | -35% | 0 | 0 | — |
case-13 | fail→pass | 3,406 | 3,503 | +3% | 1 | 1 | 0% | 518 | 946 | +83% | 0 | 0 | — |
case-14 | fail→pass | 6,531 | 2,541 | -61% | 1 | 1 | 0% | 1,206 | 761 | -37% | 0 | 0 | — |
case-15 | pass→pass | 9,997 | 5,807 | -42% | 1 | 1 | 0% | 2,112 | 1,538 | -27% | 0 | 0 | — |
case-16 | pass→pass | 6,997 | 8,684 | +24% | 1 | 1 | 0% | 1,701 | 2,301 | +35% | 0 | 0 | — |
case-17 | pass→pass | 11,090 | 9,579 | -14% | 1 | 1 | 0% | 2,314 | 2,159 | -7% | 0 | 0 | — |
case-18 | pass→pass | 8,914 | 6,023 | -32% | 1 | 1 | 0% | 1,697 | 1,369 | -19% | 0 | 0 | — |
case-19 | fail→pass | 9,480 | 2,582 | -73% | 1 | 1 | 0% | 1,595 | 771 | -52% | 0 | 0 | — |
case-20 | pass→pass | 6,948 | 1,356 | -80% | 1 | 1 | 0% | 1,173 | 538 | -54% | 0 | 0 | — |
case-21 | fail→pass | 6,431 | 1,629 | -75% | 1 | 1 | 0% | 1,200 | 604 | -50% | 0 | 0 | — |
case-22 | fail→pass | 4,352 | 1,906 | -56% | 1 | 1 | 0% | 869 | 593 | -32% | 0 | 0 | — |
case-23 | fail→pass | 3,747 | 1,759 | -53% | 1 | 1 | 0% | 720 | 611 | -15% | 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 19 counted toward the lift figure. The other 4 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 +39 percentage points is the difference between those two pass rates over the 19 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.