ricequant/ricequant-skills
11 skills11 installs286 starsGitHub
Install
npx skills add https://github.com/ricequant/ricequant-skillsSkills in this repo
1RicequantRicequant is an agent skill that turns Claude into a guided reader for the Ricequant (米筐) quantitative platform documentation. Solo builders and small quant teams use it when conversations touch RQAlphaPlus backtesting, RQData market and fundamentals pulls, factor construction with RQFactor, portfolio optimization, or performance attribution—without manually hunting through the doc site. The skill is designed to auto-trigger on Ricequant- or RQ-related queries, spin up focused retrieval against the public doc index and markdown sources, and return structured summaries the main agent can turn into config snippets, strategy hooks, or API call patterns. It matters for indie builders because quant stacks are documentation-heavy and proprietary; having procedural knowledge embedded in the agent reduces wrong-parameter backtests and speeds iteration on China-market instruments. Proprietary license terms apply per the bundled LICENSE.txt.1installs2Rq Catalyst Calendarrq-catalyst-calendar is an agent skill for solo builders and small teams who track listed equities with RiceQuant workflows. It assumes you already materialized CLI outputs—stock pool, earnings express, financials, dividends, instrument meta, announcements, and optionally curated web macro events—into a single data directory, then compiles them into a catalyst calendar report for a chosen observation interval. The output follows a fixed Chinese-language report skeleton covering executive summary, tabular calendar, breakdowns by event type, deep dives on high-impact items, recent disclosures, follow-up actions, and uncertain-date appendices. Use it when you need a repeatable, citable briefing instead of re-reading fragmented JSON during idea research, portfolio review, or ongoing watchlist maintenance. It is procedural glue around RQData contracts, not a live trading or execution integration.1installs3Rqdata Pythonrqdata-python is an agent skill for solo builders and quants who automate Chinese market research or trading analytics with Ricequant’s RQData stack inside Claude Code or similar agents. Getting symbols wrong silently poisons every downstream chart, backtest, or alert—this skill routes you through the repo’s code index manager CLI, market/type flags, and the correct contract-resolution paths for stocks, futures, and options. It encodes multi-step flows: infer underlyings (IF for CSI 300 futures, CU for copper), fetch tradable contract lists or dominant contracts, and resolve ETF option underlyings via scripted queries. A dedicated section calls out frequent API misuse—such as calling get_trading_calendar instead of get_trading_dates—and nudges you to verify codes against Ricequant conventions before get_price runs. Use it when you are building data pipelines, agent tools that answer “what is the right contract code,” or internal research notebooks that must respect exchange calendars. Complexity is advanced because domain knowledge (instrument types, underlyings) is required alongside Python execution on your machine.1installs4Rq Earnings Analysisrq-earnings-analysis packages a Ricequant-oriented earnings report workflow for solo builders and small research stacks automating equity write-ups. The SKILL content defines a full Chinese-language report skeleton—from executive summary through market expectations, management wording, balance-sheet quality, updated investment thesis, valuation, and risk—filled by scripts such as earnings-analysis/scripts/generate_report.py reading a --data-dir of JSON files. Documented inputs include company_info, industry tiers, historical_financials with revenue and profit lines, ROE history, market cap and PE/PB/dividend yield series, and price_window bars for event-window reaction. Agents use it when turning standardized quantitative extracts into citable research memos during Idea-phase diligence or when validating finance-side assumptions before building trading tools, dashboards, or content products. It is template-plus-generator rather than a live market MCP; practitioners must supply correct Ricequant exports.1installs5Rq Earnings Previewrq-earnings-preview is an agent skill for solo quant builders and indie traders who already export RiceQuant fundamentals and price history as JSON and need a repeatable earnings-week research artifact instead of ad-hoc chat summaries. Point generate_report.py at a data directory satisfying the documented contract, and the workflow stitches company metadata, industry context, multi-quarter financials, ROE history, and recent closes into one citable preview report. The template forces an executive summary, explicit forecast framing, how the name is priced versus expectations, management/announcement clues, scenario bands with plausible market reactions, and a pre-disclosure trading setup plus verification risks. It matters because margin and cash-quality rules are encoded for real-world gaps (missing gross profit on banks/insurers), so agents do not collapse entire sections to “no data.” Use it in the research phase when a disclosure date and target quarter are known and your JSON bundle is current; it does not replace live data APIs or discretionary judgment on position size.1installs6Rq Idea GenerationRQ Idea Generation is a RiceQuant agent skill for quantitative investors who need repeatable equity screening narratives instead of one-off chat opinions. Solo builders and small quant teams use it during early research to turn a defined order_book_id universe into an investment creative report dated [[REPORT_DATE]] with value, growth, and quality candidate tables plus overlap analysis. Scripts materialize idea_screening_snapshot.json from on-disk JSON inputs, then the agent enriches narrative summaries before generate_report.py renders the final Chinese Markdown artifact. It assumes you already curate stock pools and latest financials in the documented schemas, including quarter distribution and sector breakdowns. The workflow suits Claude or Cursor agents wired into a local --data-dir rather than live brokerage execution, and it pairs naturally with downstream backtest or validation skills once candidates are shortlisted.1installs7Rq Initiating Coveragerq-initiating-coverage is an agent skill for solo builders and indie analysts who need institutional-style first-coverage equity write-ups without rebuilding section order from scratch. It pairs a fixed Chinese report outline (执行摘要 through 风险提示 and appendix) with a explicit data contract that generate_report.py consumes from a local data directory. You supply normalized JSON for listing metadata, industry classification, share structure, shareholder concentration, and multi-quarter financials; the skill drives consistent narrative blocks for ownership, sell-side expectations, trading/dividend context, and comparable valuation. It suits validate-phase diligence when you are scoping whether to cover, model, or build tooling around a ticker—not live execution or portfolio ops. Complexity is intermediate because you must source or export compliant RiceQuant-shaped datasets and interpret financial quality sections yourself.1installs8Rq Morning Noterq-morning-note is an agent skill for solo quant and indie trading operators who need a repeatable 晨会纪要 (morning meeting note) without hand-assembling headlines from scattered files. It drives a report from `morning-note/scripts/generate_report.py`, reading JSON under `--data-dir` and filling templated slots such as report date, as-of time, lookback window, and coverage scope. The README defines strict consumption rules—for example, older earnings must not be labeled as overnight updates, and price-based commentary needs two observations per symbol. That discipline keeps agent-generated prose aligned with auditable inputs, which matters when you are the only analyst on the desk. Use it when you already maintain RiceQuant-style pools and nightly extracts and want Claude Code or Cursor to draft the note structure while you validate numbers. It is a structured reporting workflow, not a live market data connector.1installs9Rq Report Rendererrq-report-renderer is an agent skill for solo and indie quant builders who need consistent, professional HTML report shells instead of ad-hoc notebooks or raw tables. It encodes a RiceQuant-inspired visual system—warm paper tones, serif body type, sticky outline sidebar, and structured surfaces for tables and code—so agents can wrap backtest summaries, factor research, or portfolio analytics in a single reusable template. Use it after you have numbers or narrative copy but before you email investors, post research notes, or archive a strategy review. It is not a data pipeline or charting library; it focuses on presentation layer markup and styling conventions that make quant output readable and citable. Intermediate builders familiar with HTML/CSS and quant workflows get the most value when pairing this with their own content generation or export steps.1installs10Rq Sector Overviewrq-sector-overview is a RiceQuant-aligned agent skill for solo builders and indie quant-adjacent founders who need an industry overview memo without hand-writing every section from scratch. It expects curated JSON under a data directory—definitions, stock pools, financials, ROE, market cap, and industry maps—and drives sector-overview/scripts/generate_report.py to fill a Chinese markdown template with executive summary, sector state, financial structure, competition, valuation, relative performance, investment cues, and risk blocks. The skill is opinionated about data shape so agents do not improvise metrics or mix industry cohorts. Use it in early research when you are deciding whether a vertical is investable, crowding, or margin-compressed, and you already have or can export RiceQuant-style datasets rather than live brokerage APIs inside the agent session.1installs11Rq Thesis Trackerrq-thesis-tracker is a RiceQuant-oriented agent skill that turns a folder of standardized market and fundamentals JSON into a full investment thesis tracking report for one company. It is built for solo quants, indie research workflows, and small funds who document a core view, pillars, catalysts, and risks and want the agent to assemble executive summary, verification tables, performance review, and update logs from fresh inputs rather than rewriting Word docs. The generator scans --data-dir for files such as thesis_definition.json, latest_financials.json, price_6m.json, hs300_6m.json, pe_ratio.json, announcement_raw.json, and optional web_search_findings.json, then fills a fixed section outline from REPORT_DATE through appendix口径说明. Optional thesis_definition.json encodes thesis_name, confidence_label, target_price, planned_catalysts, and risk_items with monitor/response hooks. This is a specialized finance generator: intermediate to advanced familiarity with equity research structure and local data prep is assumed.1installs