vague

Use when writing Vague (.vague) files - a declarative language for generating realistic test data with superposition, constraints, and cross-references

⭐5stars🔀1forks📁View on GitHub🕐Updated Jan 11, 2026

When & Why to Use This Skill

This Claude skill facilitates the creation of realistic and complex test datasets using Vague, a declarative language designed for high-fidelity data generation. It solves the challenge of manual data entry and brittle mock data by allowing developers to define sophisticated schemas with built-in constraints, probabilistic values (superposition), and cross-references, ensuring that test environments accurately reflect real-world business logic.

Use Cases

  • Automated Database Seeding: Rapidly generate thousands of unique, schema-compliant records such as invoices, user profiles, or transaction logs to populate development and staging environments.
  • Edge Case and Boundary Testing: Use declarative constraints and ranges to produce specific data variations, helping QA teams identify bugs in validation logic and system stability under extreme conditions.
  • Relational Data Modeling: Create complex datasets where fields are dynamically calculated or reference other entities, such as generating orders with corresponding line items, tax calculations, and status transitions.
  • Realistic Performance Benchmarking: Generate large-scale, non-repetitive datasets that mimic production data distributions to test system performance and query optimization.
namevague
descriptionUse when writing Vague (.vague) files - a declarative language for generating realistic test data with superposition, constraints, and cross-references

Vague Language

Quick Start

schema Invoice {
  id: unique int in 1000..9999,
  status: "draft" | "sent" | "paid",
  total: decimal in 100.00..5000.00,
  line_items: 1..5 of LineItem,
  tax: round(total * 0.2, 2),
  assume total > 0
}

dataset TestData {
  invoices: 100 of Invoice
}

Core Syntax

  • Types: string, int, decimal, boolean, date
  • Superposition: "a" | "b" or weighted 0.7: "a" | 0.3: "b"
  • Ranges: int in 1..100, date in 2020..2024
  • Collections: 1..5 of Item or 100 of Item
  • Computed: total: sum(items.amount), median(), first(), last(), product()
  • Constraints: assume due_date >= issued_date
  • Refine: } refine { if type == "A" { field: int in 1..10 } }
  • References: any of companies where .active == true
  • Parent ref: = ^parent_field
  • Nullable: string? or string | null
  • Unique: id: unique int in 1..1000
  • Private: age: private int (generated but excluded from output)
  • Ordered: [48, 52, 55, 60] (cycles through values)

Reference Files