Gatedata Market Research

GateData Market Research

General Rules

⚠️ STOP — Read gatedata-runtime-rules.md first.

• MCP server: GateData (production)
• Scope required: markets
• Primary tool: market_data_query (multi-task)
• Prefer first-class task values. Legacy field_group / alias tasks (top_of_book, recent_trades, …) still work as aliases — prefer first-class names below.

When to use this skill

Equity (us_equity / hk_equity / …) this phase: snapshot, kline, orderbook/top_of_book, technical_indicators. Other surfaces are crypto-only (equity → no_data / crypto_only).

Important:

• Bare task: summary without field_group is snapshot-only — prefer first-class tasks instead of bare summary.
• task: anomaly and task: move are thin snapshot wrappers. For “why did it move?”, use kline + signals content_search/sentiment.

Do not use for prediction markets → gatedata-prediction-markets.
Do not use for news/sentiment → gatedata-signals-news.
Do not use for financial statements / earnings → gatedata-fundamentals-earnings.

Key parameters

• query — free-form symbol/asset (BTC, ETH_USDT, AAPL)
• market / context.market — required for equity (us_equity, hk_equity, kr_equity, uk_equity, jp_equity, crypto)
• venue / venue_id + market_type (spot | perp) — disambiguate listings
• timeframe / interval, from / to — kline windows. Honor explicit from/to: equity live and crypto live filter candles by that window; date-only to is treated as end-of-day bound. Prefer ISO datetimes when the user names a calendar range.
• listing_id / object_refs — preferred for exact prices
• depth, limit, include_indicators
• technical_indicators: tickers+market, and/or listing_id/listing_ids (Doris US.AAPL or Gate listing_aapl_xnas), optional analysis_as_of, part_date
• field_group — legacy only; prefer first-class task

object_refs chaining

1. First call may use query (e.g. BTC, ETH, AAPL).
2. Read metadata.object_refs from the response.
3. Pass refs in subsequent calls to avoid re-resolution and 40001 ambiguous errors.

Example chain: task: snapshot → task: kline → task: orderbook, all on market_data_query with the same object_refs.

Ambiguity handling

When the API returns 40001 or metadata.candidates:

1. If user named a venue (Binance, Gate), add venue/venue_id + market_type (spot / perp).
2. If multiple listings match, ask one clarifying question OR pick listing_id from candidates.
3. For pairs like ETH_USDT without venue, always disambiguate before order book / kline.
4. Cross-domain ambiguity (stock vs token) → gatedata-assets-resolve first.
5. If assets_resolve returns HTTP 200 with resolution_status=ambiguous and candidates — ask the user; do not treat as a tool error.

Response presentation

• Always show updated_at or candle time when quoting prices.
• Mention data_status / limitations if partial or delayed.
• Free tier sample symbols: BTC, ETH, SOL (crypto); AAPL, NVDA, TSLA (TradFi).

References

• scenarios.md — example dialogues
• troubleshooting.md — auth, scope, ambiguity