What is schema drift in Langfuse prompts?

Schema drift occurs when the structure expected by your code (function schema) differs from the structure defined in your prompt's output contract, leading to silent failures or runtime errors.

How do I validate all my prompts at once?

You can use the 'validate-all' command pointing to your prompts directory and your master schema file to perform a bulk audit of your entire project.

What specific mismatches does it detect?

It identifies missing fields, type incompatibilities (such as string vs object), and missing required properties as defined in your JSON schema.

Can I use this in my CI/CD pipeline?

Yes, the validator supports a --strict mode and returns specific exit codes (0 for success, 1 for errors), making it ideal for automated CI workflows and pre-commit hooks.

Does it support nested JSON objects?

Yes, the skill performs recursive validation to ensure that complex nested structures, arrays, and sub-objects match perfectly between the contract and the schema.

Langfuse Schema Validator

Name: Langfuse Schema Validator
Author: mberto10

bymberto10

セキュリティとテスト

Ensures consistency between Langfuse prompt output contracts and function schema definitions to prevent runtime failures.

概要

The Langfuse Schema Validator is a specialized tool designed to solve the common problem of schema drift in LLM-powered applications. It automatically verifies that the output structure described within a Langfuse prompt's output contract section aligns perfectly with the actual function or tool schema used by the application. By catching field mismatches, type inconsistencies, and missing required properties during development or CI/CD, it eliminates silent failures and ensures data integrity throughout AI pipelines.

主な機能

Recursive checking of nested object structures and data types
Automated extraction of output contracts from prompt files
Validates output contracts against function and tool schemas
Batch validation support for entire prompt directories
0 GitHub stars
Strict mode with exit codes for seamless CI/CD integration

ユースケース

Debugging unexpected data structures or type errors in LLM output nodes
Validating prompt changes before production deployment to prevent pipeline breaks
Automating schema consistency checks in pre-commit hooks for AI development

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mberto10/mberto-compound schema-validator

For use in Claude.ai and ChatGPT

Download Skill

GitHub

概要