Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user wants real-time web or news results with AI-summarized answers and citations — Grok Live Search via BlockRun. Cheapest path for "what just happened" questions where freshness beats neural-semantic ranking.
.claude/skills/blockrunai-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 43% | 0% |
Real-time web + news search with AI-summarized results and citations. PRICED PER SOURCE and expensive by default: $0.025 × max_results, +5% +$0.001 → default 10 settles $0.2635 per call. Best for fresh queries; for semantic / neural research use blockrun_exa instead.
tsblockrun_search({ body: { query: "what's the consensus on the Fed's next move", sources: ["web", "news"], max_results: 10 }})
| Field | Required | Type | Notes | |---|---|---|---| | query | yes | string | Natural-language search query | | sources | no | string] | Subset of ["web","news"]. Default: both. Does NOT multiply price. | | max_results | no | number | 1–50, default 10. Drives the price — ~$0.0263 charged per source. Pass a small number to cap spend; the gateway prices the raw value and does not floor fractions. | | from_date | no | string | YYYY-MM-DD lower bound on result date | | to_date | no | string | YYYY-MM-DD upper bound |
| User intent | sources setting | |---|---| | Breaking news / today's headlines | ["news"] | | What's the CT / KOL sentiment on X | ["x"] | | Backgrounder / explainer / docs | ["web"] | | General "find current info" question | omit — defaults to all three |
tsblockrun_search({ body: { query: "Ethereum ETF approval SEC", sources: ["news","web"], max_results: 8 } })
Cost: ~$0.2110 (8 sources: $0.025 × 8 × 1.05 + $0.001).
tsblockrun_search({ body: { query: "Solana outage today", sources: ["x"], max_results: 15 } })
Cost: ~$0.3948 (15 sources: $0.025 × 15 × 1.05 + $0.001).
tsblockrun_search({ body: { query: "Ethereum Pectra upgrade", sources: ["web","news"], from_date: "2026-02-17" }})
Cost: $0.2635 (10 sources, the default — verified live).
| Use case | Tool | |---|---| | "What's happening right now?" — freshness matters | blockrun_search | | "Find the canonical paper on X" — semantic relevance matters | blockrun_exa | | "Pull the full text of these 5 URLs" — content fetch | blockrun_exa contents | | "Cited answer to a question" | both work; blockrun_exa answer is grounded in pre-indexed corpus, blockrun_search searches the live web |
max_results: 20 with one source or three sources both charge $0.526 ($0.025 × 20 × 1.05 + $0.001 — verified live). Pass a smaller max_results to cap spend.POST /v1/search| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | fail→pass | 8,778 | 2,367 | -73% | 1 | 1 | 0% | 1,560 | 1,350 | -13% | 0 | 0 | — |
case-12 | fail→fail | 7,701 | 6,109 | -21% | 1 | 1 | 0% | 639 | 1,264 | +98% | 0 | 0 | — |
case-13 | fail→fail | 5,922 | 6,306 | +6% | 1 | 1 | 0% | 212 | 1,255 | +492% | 0 | 0 | — |
case-09 | pass→fail | 21,160 | 12,093 | -43% | 1 | 1 | 0% | 3,554 | 1,677 | -53% | 0 | 0 | — |
case-10 | fail→fail | 6,156 | 8,758 | +42% | 1 | 1 | 0% | 241 | 1,385 | +475% | 0 | 0 | — |
case-01 | fail→fail | 11,523 | 5,150 | -55% | 1 | 1 | 0% | 1,277 | 1,218 | -5% | 0 | 0 | — |
case-02 | fail→fail | 15,226 | 7,179 | -53% | 1 | 1 | 0% | 1,661 | 1,275 | -23% | 0 | 0 | — |
case-03 | fail→fail | 13,977 | 9,068 | -35% | 1 | 1 | 0% | 2,448 | 1,682 | -31% | 0 | 0 | — |
case-04 | fail→pass | 8,555 | 2,864 | -67% | 1 | 1 | 0% | 1,501 | 1,476 | -2% | 0 | 0 | — |
case-05 | fail→pass | 10,553 | 4,222 | -60% | 1 | 1 | 0% | 1,781 | 1,654 | -7% | 0 | 0 | — |
case-06 | fail→pass | 8,542 | 2,737 | -68% | 1 | 1 | 0% | 1,293 | 1,424 | +10% | 0 | 0 | — |
case-07 | pass→fail | 13,063 | 6,157 | -53% | 1 | 1 | 0% | 1,820 | 1,263 | -31% | 0 | 0 | — |
case-08 | pass→fail | 19,907 | 7,099 | -64% | 1 | 1 | 0% | 2,595 | 1,409 | -46% | 0 | 0 | — |
case-14 | pass→pass | 3,280 | 2,624 | -20% | 1 | 1 | 0% | 591 | 1,472 | +149% | 0 | 0 | — |
case-15 | pass→pass | 9,235 | 3,693 | -60% | 1 | 1 | 0% | 1,306 | 1,559 | +19% | 0 | 0 | — |
case-16 | fail→pass | 7,813 | 4,198 | -46% | 1 | 1 | 0% | 1,187 | 1,694 | +43% | 0 | 0 | — |
case-17 | fail→pass | 7,016 | 3,092 | -56% | 1 | 1 | 0% | 1,165 | 1,410 | +21% | 0 | 0 | — |
case-18 | pass→pass | 6,603 | 2,021 | -69% | 1 | 1 | 0% | 1,106 | 1,331 | +20% | 0 | 0 | — |
case-19 | fail→pass | 8,065 | 3,471 | -57% | 1 | 1 | 0% | 1,537 | 1,535 | -0% | 0 | 0 | — |
case-20 | fail→pass | 6,723 | 3,254 | -52% | 1 | 1 | 0% | 1,007 | 1,470 | +46% | 0 | 0 | — |
case-21 | pass→pass | 7,568 | 3,582 | -53% | 1 | 1 | 0% | 1,161 | 1,201 | +3% | 0 | 0 | — |
case-22 | fail→pass | 13,774 | 3,150 | -77% | 1 | 1 | 0% | 2,213 | 1,410 | -36% | 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 13 counted toward the lift figure. The other 9 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 +27 percentage points is the difference between those two pass rates over the 13 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.