Skip to content

SKILL.md

Latest commit

 

History

History
191 lines (134 loc) · 6.14 KB

SKILL.md

File metadata and controls

191 lines (134 loc) · 6.14 KB
name dot
description Query company data using Dot, an AI data analyst with access to your databases. Use when users ask about data, metrics, KPIs, reports, SQL, dashboards, or say "ask dot" or "check the numbers".
compatibility Requires dot CLI binary installed and authenticated
metadata
author version
Snowboard-Software
0.1.0

dot — AI Data Analyst CLI

dot lets you query your company's databases through Dot, an AI data analyst. Dot writes SQL, runs queries, generates charts, and explains results.

Setup

If dot is not installed or not found in PATH:

  • macOS/Linux: curl -fsSL https://app.getdot.ai/install.sh | sh
  • Windows: irm https://app.getdot.ai/install.ps1 | iex

After installing, authenticate with dot login (opens browser).

When to use

Use dot when the user asks about:

  • Data questions: "What were sales last month?", "Show me top customers"
  • Metrics and KPIs: "What's our churn rate?", "Revenue by region"
  • SQL queries: "Write a query to find...", "Check the database for..."
  • Reports: "Generate a summary of...", "Break down..."
  • Any time they say "ask dot", "check the numbers", or "query the data"

Decision: catalog vs ask

  • If the user asks WHAT data is available, run dot catalog
  • If the question is vague (no specific metric, table, or time period), run dot catalog first to understand what's available, then ask a targeted question
  • If the user asks a specific data question, go straight to dot "..."

How to use

Discover available data first

Before asking questions, run dot catalog to see what data is available:

dot catalog

This returns instantly (no LLM call) and shows:

  • Available capabilities (SQL, visualizations, scheduled reports, text analysis)
  • Custom skills configured for the org
  • Data source connections with table counts
  • Top 50 tables sorted by usage, with descriptions and column/row counts
  • External assets (Looker dashboards, etc.)

Ask questions

Run dot via Bash with the question in quotes. Set a generous timeout — Dot runs a full AI analysis pipeline (SQL generation, execution, visualization) which takes 15-60 seconds, sometimes up to 2 minutes for complex queries:

dot "What were total sales last month?"

Follow-up questions

Every response includes a chat ID. Use --chat to continue the conversation:

dot "Now break down by region" --chat cli-m1abc2d-x4y5z6

When to bypass cache

Responses are cached permanently. Use --no-cache when:

  • The question involves "today", "right now", "latest", or "current"
  • The user says "refresh", "update", or "re-run"
  • The user seems unsatisfied with a previous answer
dot "What are today's sales numbers?" --no-cache

After receiving a response

  1. Parse the output text and present the explanation to the user
  2. If a chart PNG path is shown ("Chart saved to: /tmp/dot/...png"), READ the PNG file — you have multimodal capabilities and can describe what the chart shows
  3. If a CSV path is shown and the user needs detailed analysis, read and analyze the CSV data
  4. Present suggested follow-ups if they seem relevant to the user's goal
  5. If the user wants to continue, use --chat with the chat ID from the output

Output format

The output includes:

  • Text explanation — natural language answer to the question
  • SQL query — the exact SQL that was executed
  • Data preview — first rows as CSV-like text with column stats
  • Chart — saved as PNG to /tmp/dot/<chat-id>/ (read it — you're multimodal)
  • CSV data — saved locally for further analysis
  • Dot URL — link to the full interactive analysis in the browser
  • Suggested follow-ups — use these proactively if relevant

Multi-step analysis

You can orchestrate multi-step data analysis:

  1. Run dot catalog to understand available data
  2. Ask an initial question with dot "..."
  3. Read the CSV output for deeper analysis or custom calculations
  4. Ask follow-up questions using --chat to refine results
  5. Compare results across multiple queries

Caching

  • dot "question" — cached forever until --clear-cache
  • Follow-ups with --chat are never cached (always fresh)
  • dot catalog is never cached (already fast, no LLM)

Use --no-cache to force a fresh request, or --clear-cache to wipe all cached data.

Tips for good questions

  • Start with dot catalog to understand what tables and data are available
  • Be specific: include metric names, time periods, filters
  • One question at a time works best
  • Use follow-ups (--chat) to refine rather than asking compound questions
  • If you need a chart, say "show me a chart of..." or "visualize..."

Error: command not found

If dot is not found, tell the user to install it:

curl -fsSL https://app.getdot.ai/install.sh | sh
dot login

Error: Not authenticated

If you get "Not authenticated", the user needs to log in:

dot login

If dot login doesn't work (e.g., no browser available, SSO issues), the user can create an API token manually:

  1. Open the Dot dashboard in a browser (e.g., https://app.getdot.ai or your company's Dot URL)
  2. Go to Settings > Users > API Tokens
  3. Create a new token scoped to their user
  4. Use it with: dot login --token <TOKEN> --server <SERVER_URL>

Error: Authentication failed

If you get "Authentication failed", the token may have expired. The user needs to re-login:

dot login

Error: Connection failed

If dot can't reach the server, tell the user to check their internet connection. If using a custom server, verify the URL with dot status.

Debugging

Run dot status to check who is logged in, which server is configured, and whether the token is expired.

Examples

# See what data is available
dot catalog

# Simple question
dot "What were total sales last month?"

# With follow-up
dot "Compare to the same period last year" --chat cli-m1abc2d-x4y5z6

# Chart request
dot "Show me a chart of monthly revenue trend for the past 12 months"

# Specific filters
dot "Top 10 customers by order count in Q4 2025, US only"

# Force fresh answer (bypass cache)
dot "What were total sales last month?" --no-cache