playbooks

home / skills / a5c-ai / babysitter / data-versioning-manager

data-versioning-manager skill

/plugins/babysitter/skills/babysit/process/specializations/domains/science/scientific-discovery/skills/data-versioning-manager

This skill helps manage data versions, provenance, and lineage to support reproducible research across experiments and collaborations.

npx playbooks add skill a5c-ai/babysitter --skill data-versioning-manager

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
1.1 KB
---
name: data-versioning-manager
description: Skill for managing data versions and provenance
allowed-tools:
  - Bash
  - Read
  - Write
metadata:
  specialization: scientific-discovery
  domain: science
  category: Reproducibility
  skill-id: SK-SCIDISC-025
---

# Data Versioning Manager Skill

## Purpose

Manage data versions, track provenance, and ensure data lineage for reproducible scientific research.

## Capabilities

- Version datasets
- Track data lineage
- Document transformations
- Enable rollback
- Support collaboration
- Generate provenance

## Usage Guidelines

1. Initialize versioning
2. Track data changes
3. Document transformations
4. Create snapshots
5. Manage branches
6. Export provenance

## Process Integration

Works within scientific discovery workflows for:
- Data management
- Reproducibility support
- Collaboration enabling
- Audit compliance

## Configuration

- Version control system
- Storage backends
- Metadata schemas
- Access controls

## Output Artifacts

- Version histories
- Provenance records
- Transformation logs
- Data snapshots

Overview

This skill manages data versions and documents provenance to make datasets reproducible and auditable. It provides deterministic, resumable controls for dataset snapshots, branching, and rollback. The goal is to preserve lineage and transformations so teams can reproduce results and meet compliance requirements.

How this skill works

The skill records every change as a versioned snapshot and stores metadata that describes source, parameters, and transformation steps. It links versions into a lineage graph so you can trace origin and dependencies across branches. Rollback, export of provenance records, and access control are supported to enforce reproducibility and collaboration policies.

When to use it

  • When you need reproducible study results or experiments
  • When multiple contributors modify shared datasets
  • When audit trails or regulatory compliance require provenance
  • When you need safe rollback after a faulty transformation
  • When preparing datasets for publication or peer review

Best practices

  • Initialize versioning early in a project to capture original sources and initial preprocessing steps
  • Use clear, machine-readable metadata schemas for transformations and parameters
  • Create frequent snapshots around major processing steps or experiments
  • Manage branches for exploratory transformations and merge only validated versions into mainline
  • Automate provenance export and storage with your existing storage backends and access controls

Example use cases

  • Track raw-to-cleaned dataset lineage in a multi-step processing pipeline
  • Maintain branches for alternative preprocessing strategies and compare outcomes
  • Generate provenance bundles for peer review or regulatory submission
  • Rollback to a prior snapshot after detecting a data-quality issue
  • Coordinate dataset updates across a distributed research team with enforced access policies

FAQ

How does rollback work?

Rollback restores a selected snapshot as the active dataset and records the action in the provenance log so the change is traceable.

What metadata should I capture?

Capture source identifiers, timestamps, tool versions, transformation parameters, responsible actors, and checksums to ensure integrity and traceability.