Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user asks "what's hot", "what's moving", "any alpha", "show me squeeze setups", "what's the setup on ETH", "is SOL coiled", "should I deploy NEAR" or any market-scan / single-pair-drilldown question. Surfaces Superior Trade's live multi-bucket scoring across Hyperliquid alts + HIP-3 (stocks/indices/commodities/FX) — Squeeze fuel, Stealth accumulation, Coiled spring, Basis flipping. The engine picks the strongest timeframe (15m/1h/4h/24h) per pair per bucket; you don't pick one. Pair
.claude/skills/superior-trade-intelligence/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 261% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -19% | 0% |
Live ranked alpha scan over Hyperliquid alts + HIP-3 markets. Returns the same Intelligence data from Superior Trade — bucket fits, per-pair best timeframes, snapshots, and recommended deploy templates.
| File | What it covers | |---|---| | references/buckets.md | The 4 buckets: Squeeze fuel, Stealth accumulation, Coiled spring, Basis flipping. Setup / Edge / Scoring + AI Critic concerns per bucket. Read this before presenting any scan output to the user. | | references/api.md | The two endpoints: GET /v2/intelligence/scan (list mode) and GET /v2/intelligence/setup/{pair} (single-pair detail). Response schemas + examples. | | references/workflow.md | Recommended end-to-end recipes: scan → pick → setup → backtest → deploy. How to translate a best_fit tuple into a /v2/backtesting call. | | references/glossary.md | Plain-English definitions for the trading terms in scan responses (funding, OI, basis, fees-paid notional, OI turnover, CVD, etc.). Reference this when explaining results to a non-trader. |
GET /v2/intelligence/scan.Pass category=both for any open-ended question. The parameter defaults to alts, so a bare call silently returns crypto perps only and never surfaces a single HIP-3 stock, index or commodity — even though this skill advertises them. Narrow to alts or tradfi only when the user asked for one. Add bucket when they named a specific setup.
GET /v2/intelligence/setup/{pair}. Always do this BEFORE backtesting / deploying so the choice is grounded in current data.Both endpoints return live, ranked data. Never substitute a market scan from training data — prices are stale, the ranking framework is Superior's, and timeframes are picked by the engine. When presenting:
best_fit.bucket_title @ best_fit.timeframe (score) tuple per pair.pct_change_24h + funding_paid_notional_usd_per_yr).bucket_fits only if asked why another bucket wasn't picked.references/buckets.md before recommending a deploy.xyz:NVDA, xyz:AAPL, etc.) trade 24/7 but the underlying equities only trade during NYSE hours — see the US Market Closed warning in references/buckets.md § Stocks.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 12,916 | 8,945 | -31% | 1 | 1 | 0% | 1,801 | 2,133 | +18% | 0 | 0 | — |
case-01 | fail→fail | 56,034 | 83,141 | +48% | 1 | 1 | 0% | 3,713 | 3,514 | -5% | 0 | 0 | — |
case-02 | fail→fail | 11,959 | 40,998 | +243% | 1 | 1 | 0% | 1,573 | 1,687 | +7% | 0 | 0 | — |
case-03 | fail→fail | 15,947 | 40,190 | +152% | 1 | 1 | 0% | 2,322 | 1,374 | -41% | 0 | 0 | — |
case-04 | fail→pass | 15,954 | 5,859 | -63% | 1 | 1 | 0% | 2,326 | 1,674 | -28% | 0 | 0 | — |
case-05 | fail→pass | 26,724 | 53,915 | +102% | 1 | 1 | 0% | 4,608 | 4,732 | +3% | 0 | 0 | — |
case-06 | fail→pass | 33,748 | 7,640 | -77% | 1 | 1 | 0% | 515 | 1,858 | +261% | 0 | 0 | — |
case-07 | fail→fail | 44,027 | 8,398 | -81% | 1 | 1 | 0% | 2,441 | 1,359 | -44% | 0 | 0 | — |
case-08 | fail→fail | 30,235 | 10,383 | -66% | 1 | 1 | 0% | 4,545 | 1,281 | -72% | 0 | 0 | — |
case-09 | fail→fail | 11,610 | 6,427 | -45% | 1 | 1 | 0% | 1,716 | 1,804 | +5% | 0 | 0 | — |
case-10 | fail→pass | 11,763 | 4,069 | -65% | 1 | 1 | 0% | 1,679 | 1,367 | -19% | 0 | 0 | — |
case-11 | fail→pass | 12,222 | 11,446 | -6% | 1 | 1 | 0% | 1,668 | 2,630 | +58% | 0 | 0 | — |
case-13 | pass→pass | 18,811 | 15,894 | -16% | 1 | 1 | 0% | 2,585 | 2,973 | +15% | 0 | 0 | — |
case-14 | fail→pass | 12,451 | 5,010 | -60% | 1 | 1 | 0% | 1,660 | 1,423 | -14% | 0 | 0 | — |
case-15 | fail→pass | 19,336 | 19,273 | -0% | 1 | 1 | 0% | 2,717 | 2,525 | -7% | 0 | 0 | — |
case-16 | fail→pass | 10,120 | 14,429 | +43% | 1 | 1 | 0% | 1,481 | 2,976 | +101% | 0 | 0 | — |
case-17 | pass→pass | 13,295 | 5,253 | -60% | 1 | 1 | 0% | 1,940 | 1,405 | -28% | 0 | 0 | — |
case-18 | pass→pass | 10,086 | 4,113 | -59% | 1 | 1 | 0% | 1,532 | 1,371 | -11% | 0 | 0 | — |
case-19 | pass→pass | 16,939 | 6,227 | -63% | 1 | 1 | 0% | 2,553 | 1,770 | -31% | 0 | 0 | — |
case-20 | fail→fail | 17,048 | 5,448 | -68% | 1 | 1 | 0% | 2,754 | 1,716 | -38% | 0 | 0 | — |
case-21 | pass→pass | 16,258 | 16,262 | +0% | 1 | 1 | 0% | 2,146 | 2,969 | +38% | 0 | 0 | — |
case-22 | pass→pass | 16,160 | 10,114 | -37% | 1 | 1 | 0% | 2,239 | 2,160 | -4% | 0 | 0 | — |
case-23 | fail→fail | 14,862 | 39,625 | +167% | 1 | 1 | 0% | 2,568 | 1,066 | -58% | 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 17 counted toward the lift figure. The other 6 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 17 comparable cases. 1 case got worse with the skill loaded, and it is 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.