sft-data-format

Validate JSON Lines SFT data formats, idx resume keys, and metadata structures.

Updated Jan 16, 2026
One-click install
npx skills add https://github.com/HsunGong/prep --skill sft-data-format
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sft-data-format
Source: https://github.com/HsunGong/prep/tree/main/.github/skills/data-format
Command: npx skills add https://github.com/HsunGong/prep --skill sft-data-format

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps teams validate and document data formats and metadata structures used in the SFT pipeline, reducing integration errors and improving reproducibility.

Core Features & Use Cases

  • Data format validation: Ensures input/output data follow the JSON Lines convention and the stage metadata structure.
  • Resume key handling: Verifies usage of idx as a resuming key across pipeline steps.
  • Metadata documentation: Produces clear metadata schemas and examples for downstream components.
  • Use Case: When ingesting datasets into the SFT pipeline, run this skill to confirm formatting, keys, and metadata are consistent before processing.

Quick Start

Use the sft-data-format skill to validate your dataset's JSON Lines formatting, idx resume keys, and think tags.

Frequently Asked Questions about sft-data-format

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate JSON Lines formatting for an SFT data pipeline?

You validate JSON Lines formatting by checking that each row follows the convention and includes proper idx resume keys and think tags. This skill enforces consistent input/output formats and standardized metadata structures across pipeline stages before processing.

What is an idx resume key and when do I need it in SFT data ingestion?

An idx resume key is a resuming key used across pipeline steps to track data position. You need it during SFT data ingestion and preprocessing to verify proper resume key handling and ensure reproducible, consistent pipeline metadata.

How do I document metadata schemas for downstream components in a model training workflow?

You document metadata schemas by producing clear schema definitions and examples for downstream components. This skill generates standardized metadata representations that clarify data formats and reduce integration errors in model-training workflows.

Does this data format validation approach work with think tags in JSONL files?

Yes, the validation approach works with think tags in JSONL files by enforcing consistent input/output formats and proper tag handling. It verifies that think tags and idx resume keys follow the standardized metadata structure across pipeline stages.

What's the best way to prevent integration errors when ingesting datasets into an SFT pipeline?

The best way to prevent integration errors is to run format validation before processing to confirm JSON Lines formatting, idx resume keys, and metadata are consistent. This reduces integration errors and improves reproducibility across the SFT pipeline.

Why does my SFT pipeline fail when metadata structures are inconsistent across stages?

SFT pipelines fail from inconsistent metadata structures because downstream components expect standardized schemas and proper idx resume key handling. Validating JSON Lines formatting and metadata consistency before processing prevents these integration errors.