What problem does it solve? Futures DeepView data spans 35+ interfaces with mixed conventions and large single-day snapshots, so agents calling them raw tend to invent parameters, stitch cross-day data into fake signals, and present inference as fact. This Skill turns natural-language requests like "analyze broker-seat battles in rebar" into disciplined Pandadata call plans and produces Chinese research reports with facts and inference strictly separated. ## Core Features & Use Cases - Four analysis modes: seat battle (net position margin, build process, P&L ranks), structure analysis (basis percentile, term structure, calendar spreads), inventory & spot (warehouse receipts, virtual ratio, spot profit), and all-variety arbitrage scanning. - Built-in guardrails: request triage, contract verification via the pandadata-api skill, small-sample trial calls, same trade_date alignment across modules, and evidence-chain annotation on every conclusion. - Structured reporting: a fixed 9-section Chinese report skeleton (summary, data scope, quotes, seat battle, basis & term structure, inventory, cross-validation, risks, appendix) or a compressed 4-part quick answer. - Use Case: Ask "分析一下螺纹钢最近20个交易日的席位博弈" and receive a report combining price trend, broker net position margin, build process, and profit rankings with method names, parameters, and data cutoff dates cited. ## Quick Start Ask the agent to analyze the broker position battle, basis, and warehouse receipts for the rebar dominant contract over the last 20 trading days.
