Save Hubble Query to Obsidian

Save Hubble queries to ~/Documents/stripe-obsidian/Queries/<project>/ with rich metadata for efficient searching later.

Workflow

1. Get Hubble URL and Project

Required inputs:

•Hubble query URL (e.g., https://hubble.corp.stripe.com/queries/mingc/6485adb6)
•Project name (e.g., payments-intelligence, risk, data-platform)

If project not provided: Ask user which project this belongs to.

Extract from URL:

•Query ID: Last part of URL path (e.g., 6485adb6)
•Author: Username in URL path (e.g., mingc)

2. Fetch Query Content

Use Hubble MCP tool to get query metadata and SQL:

code

get_hubble_query_metadata(permalink="<hubble_url>")

This returns:

•Query title
•Description (if available)
•SQL query text
•Column types
•Execution metadata

3. Parse SQL for Metadata

Analyze the SQL query text to extract:

Input tables: Look for table references in FROM and JOIN clauses

•Pattern: fact.charges, dim.merchants, analytics.daily_summary
•Extract schema and table name

Output columns: Look at SELECT clause

•Extract column names and aliases
•Note aggregations (SUM, COUNT, AVG, etc.)

Query complexity: Assess based on:

•low: Simple SELECT with basic WHERE clause, 1-2 tables
•medium: JOINs, aggregations, GROUP BY, 3-5 tables
•high: Subqueries, CTEs, window functions, 5+ tables

Query type: Infer from SQL patterns

•analytical: Aggregations, GROUP BY, business metrics
•experiment: Tables/columns containing "exp", "test", "variant"
•debugging: LIMIT clauses, specific IDs in WHERE
•monitoring: Time-based filters, recent data, counts

4. Generate Filename

Create semantic filename from query content:

Rules:

•Use kebab-case: smart-retry-model-score-distribution
•Base on: query title > main operation > primary table
•Keep it descriptive but concise (3-7 words)
•Avoid generic terms like "query" or "analysis" unless specific

Examples:

•"Smart Retry Model Score Distribution" → smart-retry-model-score-distribution.md
•"Payment Success Rate by Country" → payment-success-rate-by-country.md
•"Chargeback Trends Last 90 Days" → chargeback-trends-90-days.md

5. Suggest Tags

Based on SQL analysis, suggest relevant tags:

Data source tags:

•Tables: #fact-charges, #dim-merchants
•Systems: #presto, #iceberg

Operation tags:

•#aggregation, #join, #window-function

Domain tags:

•#payments, #fraud, #ml-model, #risk, #experiment

Purpose tags:

•#debugging, #monitoring, #analysis, #experiment

Present to user: "Suggested tags: experiment, smart-retry, ml-model, score-analysis. Any changes?"

6. Create Markdown File

File path: ~/Documents/stripe-obsidian/Queries/<project>/<filename>.md

File structure:

markdown

---
id: <query_id>
title: <query_title>
author: <author_username>
created: <current_timestamp_iso8601>
project: <project_name>
tags: [<tag1>, <tag2>, <tag3>]
tables:
  - <schema.table1>
  - <schema.table2>
output_columns:
  - <column1>
  - <column2>
hubble_url: <full_url>
visualization: <chart_type_if_available>
query_type: <analytical|experiment|debugging|monitoring>
complexity: <low|medium|high>
---

## Description

<query_description_or_generated_summary>

## SQL Query

```sql
<sql_query_text>

Notes

<optional_user_notes>

code


### 7. Confirm and Save

**Create project directory if needed:**

```bash
mkdir -p ~/Documents/stripe-obsidian/Queries/<project>

Use Write tool to create the file:

code

Write(
  file_path="~/Documents/stripe-obsidian/Queries/<project>/<filename>.md",
  content=<generated_markdown>
)

Confirm to user:

•File location
•Filename generated
•Tags applied
•How to find it later with find-query skill

Example

User: "Save this query to payments-intelligence: https://hubble.corp.stripe.com/queries/mingc/6485adb6"

Actions:

•Extract: ID=6485adb6, author=mingc, project=payments-intelligence
•Fetch query metadata via Hubble MCP
•
Parse SQL finds:
- •Tables: fact.charges, exp.smart_retry_scores
- •Columns: score_bucket, charge_count, conversion_rate
- •Type: experiment (has "exp" table)
- •Complexity: medium (JOIN + aggregation)
•Generate filename: smart-retry-model-score-distribution.md
•Suggest tags: experiment, smart-retry, ml-model, score-analysis
•Create file with complete metadata
•Confirm: "✅ Saved to Queries/payments-intelligence/smart-retry-model-score-distribution.md"

Edge Cases

Query has no title:

•Generate from first few words of SQL or main operation
•Example: "SELECT merchant_id..." → merchant-payment-summary

Tables can't be parsed:

•Mark as tables: [] and add note to review manually
•Still save the query but flag incomplete metadata

Project name unclear:

•Ask user explicitly: "Which project? (payments-intelligence, risk, data-platform, etc.)"
•Suggest based on table names if possible

Duplicate filename:

•Append date suffix: query-name-2026-02-09.md
•Or ask user for custom name

Very long SQL:

•Still save everything
•Add note that query is complex and may need documentation

Tips

•Extract as much metadata as possible automatically
•Keep filename semantic and searchable
•Be generous with tags - better to over-tag than under-tag
•If user provides additional context, add to Notes section
•Preserve SQL formatting from Hubble