MarkBack

Appearance

MarkBackMarkdown-compatible labels for content + feedback

A compact, lintable format and toolchain for pairing content with single-line feedback in LLM prompt workflows.

Get Started
Format Spec
GitHub
PyPI
npm

Format-First

Records are easy to read, easy to lint, and stay stable under diff.

Storage Flexibility

Single files, multi-record files, compact label lists, and paired files.

Tooling Included

CLI and Python APIs for parsing, linting, and normalization.

What MarkBack Is ​

MarkBack is a Markdown-based format and toolchain for labeling content with feedback in a single, lintable line. It keeps raw content and evaluation notes together so you can iterate prompts, train evaluators, and track revisions without inventing a custom schema.

Quick Example ​

mb
@uri local:prompt-001

# API Design Task

Create an endpoint that:

- **GET** `/users/{id}`
- Returns JSON with `name`, `email`
- Handle 404 for missing users
<<< approved; clarity=high; needs=error examples

The content (Markdown) stays readable. The <<< line captures feedback. Try it live →

What Problem It Solves ​

Prompt iteration usually scatters content, labels, and reviewer notes across JSON, spreadsheets, and ad-hoc scripts. MarkBack standardizes that feedback loop in a text format that works in Git, can be linted, and scales from a single prompt to thousands of records.

Conceptual Flow ​

content + feedback  ->  MarkBack records  ->  lint / normalize  ->  prompt iteration