Agent Skills
› clidey/whodb
› query-builder
query-builder
GitHub将自然语言数据查询转换为SQL。通过解析用户意图，结合数据库Schema生成并执行SQL查询，支持多数据库方言及安全限制，适用于自助数据分析场景。
Trigger Scenarios
用户用自然语言询问数据统计或筛选条件
需要基于当前数据库结构生成查询语句
Install
npx skills add clidey/whodb --skill query-builder -g -y
SKILL.md
Frontmatter
{
"name": "query-builder",
"description": "Convert natural language questions into SQL queries. Activates when users ask data questions in plain English like \"show me users who signed up last week\" or \"find orders over $100\"."
}
Query Builder
Convert natural language questions into SQL queries using the database schema.
When to Use
Activate when user asks questions like:
- "Show me all users who signed up last month"
- "Find orders greater than $100"
- "Which products have low inventory?"
- "Get the top 10 customers by total spend"
Workflow
1. Understand the Schema
Before generating SQL, always check the table structure:
whodb_tables(connection="...") → Get available tables
whodb_columns(table="relevant_table") → Get column names and types
2. Identify Intent
Parse the natural language request:
- Subject: What entity? (users, orders, products)
- Filter: What conditions? (last month, > $100, active)
- Aggregation: Count, sum, average, max, min?
- Grouping: By what dimension?
- Ordering: Sort by what? Ascending/descending?
- Limit: How many results?
3. Map to Schema
- Match entities to table names
- Match attributes to column names
- Identify foreign key joins needed
4. Generate SQL
Build the query following SQL best practices:
SELECT columns
FROM table
[JOIN other_table ON condition]
WHERE filters
[GROUP BY columns]
[HAVING aggregate_condition]
ORDER BY column [ASC|DESC]
LIMIT n;
5. Execute and Present
whodb_query(query="generated SQL")
Translation Patterns
| Natural Language | SQL Pattern |
|---|---|
| "last week/month/year" | WHERE date_col >= DATE_SUB(NOW(), INTERVAL 1 WEEK) |
| "more than X" / "greater than X" | WHERE col > X |
| "top N" | ORDER BY col DESC LIMIT N |
| "how many" | SELECT COUNT(*) |
| "total" / "sum of" | SELECT SUM(col) |
| "average" | SELECT AVG(col) |
| "for each" / "by" | GROUP BY col |
| "between X and Y" | WHERE col BETWEEN X AND Y |
| "contains" / "like" | WHERE col LIKE '%term%' |
| "starts with" | WHERE col LIKE 'term%' |
| "is empty" / "is null" | WHERE col IS NULL |
| "is not empty" | WHERE col IS NOT NULL |
Date Handling by Database
PostgreSQL
WHERE created_at >= NOW() - INTERVAL '7 days'
WHERE created_at >= DATE_TRUNC('month', CURRENT_DATE)
MySQL
WHERE created_at >= DATE_SUB(NOW(), INTERVAL 7 DAY)
WHERE created_at >= DATE_FORMAT(NOW(), '%Y-%m-01')
SQLite
WHERE created_at >= DATE('now', '-7 days')
WHERE created_at >= DATE('now', 'start of month')
Examples
"Show me users who signed up this month"
SELECT * FROM users
WHERE created_at >= DATE_TRUNC('month', CURRENT_DATE)
ORDER BY created_at DESC;
"Find the top 5 products by sales"
SELECT p.name, SUM(oi.quantity) as total_sold
FROM products p
JOIN order_items oi ON p.id = oi.product_id
GROUP BY p.id, p.name
ORDER BY total_sold DESC
LIMIT 5;
"How many orders per customer?"
SELECT customer_id, COUNT(*) as order_count
FROM orders
GROUP BY customer_id
ORDER BY order_count DESC;
Safety Rules
- Always use LIMIT for exploratory queries (default: 100)
- Never generate DELETE, UPDATE, or DROP unless explicitly requested
- Warn if query might return large result sets
- Use table aliases for readability in JOINs
Version History
- 1ed87bc Current 2026-08-20 10:41
