Skip to content

fd-daas-research

Orchestrate a full DAAS research workflow. This is the flagship skill - it turns a natural-language demand into a persisted study with indicators, a dashboard, and a markdown report.

When it triggers

Phrases like:

  • "帮我研究一下比亚迪，做指标和看板"
  • "research TSLA - build indicators and a dashboard"
  • "研究一下沪深300成分股做个看板"
  • "research SSE banks - set up indicators and a dashboard"
  • "分析这只股票并做个可视化"

...or any entity + "research / 分析 / 研究 / full pipeline".

The flow

analyze demand -> [entity collection] -> indicators -> dashboard
              -> persist as `research` bundle -> generate markdown report
  1. Analyze the demand into a plan (uses sqlite3 for plan context).
  2. Build components via the fd-daas-* creator skills:
  3. If the demand names a group (watchlist, a rule like "SSE banks", an explicit list of codes), it first delegates to fd-daas-entities-collection-creator to persist the group.
  4. Builds indicators over the members.
  5. Builds a dashboard.
  6. Persist the whole study as a researches row using the research_* MCP tools (research_create).
  7. Generate a markdown report (research_generate_report) written to researches/<name>.md.
  8. Emits a skill-run-notification block.

Refresh vs. rebuild

On a repeat demand whose research already exists, this skill refreshes it (research_refresh - recompute indicators, re-sync rule-based collections) and regenerates the report instead of rebuilding from scratch. It auto-detects prior researches.

What it does NOT do

  • Only indicators? Use fd-daas-indicators-creator.
  • Only a dashboard? Use fd-daas-dashboard-creator.
  • A one-shot fetch? Use fd-daas-fetch-data.
  • Only an entity collection? Use fd-daas-entities-collection-creator.

This skill orchestrates the full analyze -> [collection] -> indicators -> dashboard -> research-bundle flow.

Example

Research BYD's trend: build the entity + indicator collections, compute the indicators, make a dashboard, and persist it as a research bundle with a markdown report.

Result:

researches/byd-trend.md          # generated report
researches row in daas.db        # persisted bundle
dashboards/<slug>/index.html     # dashboard

MCP equivalents

research_create(name="byd-trend", entity_collection_name=..., indicator_collection_name=..., dashboard_slug=..., pipeline_collection_name=...)
research_generate_report(name="byd-trend")
research_refresh(name="byd-trend")        # later, to refresh data
research_get(name="byd-trend")            # inspect

Works for any scope

Company, industry (a collection of stocks), city, or country - the collections you build define the scope. See Create a Research.