AgentSkillsCN

create-dashboard

在 Python 中，借助 Durable Task SDK 与 Azure Durable Task Scheduler，构建持久、容错的工作流。适用于编排、活动、实体的创建，或函数链式调用、扇出/扇入、人工交互，以及有状态代理等模式的实现。无论您的 Python 应用是否依赖 Azure Functions，此技能都能满足持久执行、状态持久化，或分布式事务的需求。

SKILL.md
--- frontmatter
name: create-dashboard
description: Builds comprehensive Amplitude dashboards from requirements or goals, organizing charts into logical sections with appropriate layouts. Use when creating a complete dashboard from scratch or assembling existing charts into a cohesive view.

Create Dashboard

Create new team or initiative dashboards, organize scattered charts, build executive reporting, or set up review cadence dashboards.

Instructions

Step 0: Discovery (if unfamiliar with the feature)

Before building, understand what you're tracking:

  • •Search for existing dashboards/charts related to the topic
  • •Search for relevant events: Amplitude:search with entityTypes: ["EVENT", "CUSTOM_EVENT"]
  • •Use get_event_properties to understand available properties for segmentation
  • •Ask user for clarification on primary goals, key segments, or time horizons

Step 1: Define Dashboard Purpose

Clarify:

  • •Who is the audience?
  • •What decisions will it inform?
  • •How frequently will it be reviewed?
  • •What's the narrative structure?

Step 2: Gather or Create Charts

If existing charts found (>5 relevant):

  • •Use Amplitude:search to find relevant existing charts
  • •Use Amplitude:get_charts to retrieve their definitions
  • •Identify gaps that need new charts

If few/no charts exist (<5 relevant):

  • •Switch to "greenfield build" mode
  • •Use Amplitude:query_dataset to create needed charts
  • •Save all charts with Amplitude:save_chart_edits before building dashboard
  • •Consider searching for relevant events first with entityTypes: ["EVENT", "CUSTOM_EVENT"]

Creating new charts:

  • •Prototype with query_dataset to verify data
  • •Save in batches using save_chart_edits (more efficient)
  • •Collect all chart IDs before creating dashboard

Step 3: Plan the Layout

Organize into logical sections:

  1. •Summary Row: Key metrics at a glance (headline view)
  2. •Trend Section: How things are changing
  3. •Breakdown Section: Segments and dimensions
  4. •Detail Section: Supporting analyses

Step 4: Create the Dashboard

Use Amplitude:create_dashboard with:

  • •Clear, descriptive name
  • •Rows with appropriate heights (375, 500, 625, or 750px)
  • •Charts sized appropriately (3-12 columns)
  • •Rich text headers for sections
  • •Chart display configurations in chartMetas:
    • •metric_only: Headline KPIs (single number)
    • •series: Trend lines (default view)
    • •converted: Funnels (conversion view)
    • •table: Data tables

Step 5: Add Context

Include rich text blocks for:

  • •Dashboard purpose and audience
  • •How to interpret key metrics
  • •Links to related resources
  • •Last updated or review schedule

Layout Guidelines

Content TypeSuggested WidthSuggested Height
Headline metric3-4 columns375px
Trend chart6-12 columns500px
Comparison chart6 columns500px
Detailed table12 columns625px
Section header12 columns375px

Best Practices

  • •Put most important metrics above the fold
  • •Use consistent chart sizing within rows
  • •Group related metrics together
  • •Add explanatory text for complex metrics
  • •Ask user about focus areas if multiple valid approaches exist
  • •Create charts in batches to minimize tool calls

Common Issues

Query errors (500/400):

  • •Simplify: remove complex groupBy, reduce date ranges, avoid nested properties
  • •Verify events/properties exist using search first
  • •Use eventsSegmentation with groupBy instead of dataTableV2 for top N lists

No data returned:

  • •Check event names are exact matches (case-sensitive)
  • •Verify date range covers when events were tracked
  • •Confirm user segments aren't too restrictive