Loading ToolBench
Read-only crypto market research, Hyperliquid and cross-exchange spread tools from Market Depth Lab.
Strong schema definitions with proper enums, ranges, and descriptions across all 9 tools. Tool names follow verb_noun pattern (calculate_, get_, analyze_, list_, scan_). Descriptions are detailed (100-250 chars typical) and explain purpose, data sources, and limitations. All parameters have types and descriptions. Output schemas documented with structured envelopes. Key weakness: descriptions lack explicit 'when to use' guidance and some parameter descriptions could be more prescriptive about constraints. Error handling present but recovery guidance minimal. No security/permission documentation despite read-only nature.
Read the public main-account state for one Hyperliquid address: perpetual positions, spot balances, open orders, recent fills, exposure, leverage, PnL and fee totals. No private key, signature or wallet connection is requested.
Estimate Hyperliquid spot or perpetual trading fees from notional, maker/taker role, fee tier, discounts, asset multiplier, maker rebate, and builder fee. This is a transparent estimate, not an exchange quote.
Fetch a live HYPE spot/perpetual snapshot from Hyperliquid and convert either HYPE to USD or USD to HYPE. Reports spot-perp basis, 24h spot change, and 24h spot notional when available.
List publicly visible active Hyperliquid TWAP schedules and estimate current notional using live market references. Schedule time is known; execution progress and unfilled quantity are not inferred.
List current perpetual funding rates on Hyperliquid main and HIP3 DEXs. Rates are normalized to hourly decimal format. Results include mark price, open interest, 24h volume and next settlement time.
Parameter descriptions lack explicit constraint guidance. E.g., 'Staking discount percentage' does not state '0-40%' in text; constraint is only in schema. LLMs cannot reliably read JSON Schema, they need constraints in description text.
Tool descriptions do not include 'when to use' guidance or differentiation from similar tools. E.g., list_large_hyperliquid_positions vs list_top_hyperliquid_holders, when should an agent choose one over the other? Descriptions lack this decision logic.
No explicit error handling or recovery guidance in tool descriptions. E.g., 'analyze_hyperliquid_wallet' does not state what happens if address is invalid or not found, or what the agent should do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2025-06-18+ | v2 |
Screen the monitored-address cohort for large perpetual positions by notional, leverage, or PnL. Results are cohort-relative, not protocol-wide. Positions are ranked within each market.
Screen the monitored-address cohort for top holders of a specific perpetual or spot asset. Results are cohort-relative, not protocol-wide. Holders are ranked by position size.
Compare current perpetual funding on Binance USD-M, OKX and Hyperliquid. For each symbol, the scanner pairs a long on the lower hourly funding rate with a short on the higher rate, normalizes different settlement intervals, and estimates gross carry, user-supplied costs and break-even time. It does not place orders or describe the result as risk-free arbitrage.
Compare best bid and ask quotes for selected perpetual symbols on Binance USD-M, OKX and Hyperliquid. The engine ranks the cheapest cross-venue buy against the richest cross-venue sell, reports top-of-book capacity and subtracts configurable taker-fee estimates. It is a persistent-divergence screen, not a latency-arbitrage or order-execution tool.
Descriptions include risk disclaimers (e.g., 'not risk-free arbitrage', 'not an execution quote') but these are buried in methodology/risk_boundary fields in response, not in tool description. LLMs may not read response structure before calling the tool.
No permission or scope declarations. Tools are read-only but this is not formally declared in tool metadata. No audit trail or logging configuration visible in source.