AgentSkillsCN

ask-statistician

向Varro AI统计师提问有关丹麦统计与数据的问题。当需要查询丹麦统计数据、创建仪表板或发布仪表板时使用。支持多轮对话，并可指定--chat-id。

SKILL.md
--- frontmatter
name: ask-statistician
description: >
  Ask the Varro AI statistician questions about Danish statistics and data.
  Use when you need to query Denmark's statistical data, create dashboards,
  or publish dashboards. Supports multi-turn conversations with --chat-id.

Ask Statistician

Send questions to the AI statistician and get text answers back.

Ask a question

bash
uv run python .claude/skills/ask-statistician/scripts/ask.py "What is the population of Denmark?"

Resume a conversation (multi-turn):

bash
uv run python .claude/skills/ask-statistician/scripts/ask.py --chat-id 42 "Break that down by region"

Output format (one line per field, parseable):

code
chat_id: 42
current_url: /dashboard/befolkning
response: The population of Denmark is approximately 5.9 million...

Save the chat_id from the first call to reuse in follow-ups. The current_url tracks what dashboard the statistician navigated to.

Create a dashboard

Ask the statistician to create a dashboard. Be directive about structure:

bash
uv run python .claude/skills/ask-statistician/scripts/ask.py "Create a dashboard showing housing prices by region since 2010"

The statistician will create the dashboard files and navigate to it. The current_url in the output tells you the dashboard path.

Publish a dashboard

After the statistician creates a dashboard, publish it to get a public URL:

bash
uv run python .claude/skills/ask-statistician/scripts/publish.py --slug boligpriser

Output:

code
url: https://varro.dk/public/1/boligpriser

Options

Both scripts accept --user-id N (default: 1).

ask.py also accepts:

  • •--chat-id N — resume an existing chat (omit to create new)
  • •--current-url /path — set the initial dashboard URL context

Shell state

Each ask.py invocation starts a fresh Jupyter shell. The statistician re-queries data as needed. Message history persists across turns via --chat-id, so the agent has full conversational context.

Timeouts

Agent runs can take 30-120 seconds for complex queries. Use a generous timeout (e.g. 300000ms) when calling the scripts.