Dataset Schema Designer Skill

Purpose

Define structured, FAIR-compliant, and standards-driven dataset schemas for research projects. Explicitly surface required/optional fields, metadata, provenance, versioning, schema risks, harmonization considerations, downstream analysis implications, and validation rules. Produce machine-readable QA Validation Block for downstream skills.

Scope

Human, animal, or simulation datasets
Outputs field definitions, types, metadata, provenance, versioning, and validation rules
Upstream: Protocol Writer (Protocol Handoff Block)
Downstream: Dataset QA, Feature Engineering

When to Use

Before data collection, during protocol development, or for schema review/harmonization, with Protocol Handoff Block available

When Not to Use

If research goals or data collection plan are missing or ambiguous

Required Inputs

Protocol Handoff Block from Protocol Writer
Data types and sources

Optional Inputs

Domain-specific metadata requirements (e.g., MIAME, BIDS)
Output format preference (Markdown, table)

Workflow

Parse Protocol Handoff Block, data collection plan, and data types
Propose dataset fields, types, value ranges, and validation rules
Define metadata (units, standards, provenance, versioning) and missing data policy
Identify schema risks and harmonization considerations
Surface downstream analysis implications
Output structured schema with explicit rationale for field choices
Generate QA Validation Block for downstream skills (machine-readable)

Output Contract

Field Table (name, type, description, units, range, required/optional, validation rules)
Metadata Section (standards, provenance, harmonization notes, versioning)
Missing Data Policy (handling, imputation, reporting)
Schema Risks and Harmonization Considerations
Downstream Analysis Implications
QA Validation Block (machine-readable)
Errors (if any)

Quality Checks

All fields are defined, described, and justified
Metadata, provenance, and versioning are explicit
Schema risks and harmonization are surfaced
Validation rules are present and actionable
Output is suitable for direct use by Dataset QA and Feature Engineering skills

Failure Modes

Missing, ambiguous, or unjustified field definitions
Non-compliance with standards or FAIR principles
Failure to address missing data, harmonization, or validation
Lack of downstream analysis implications

Domain Adaptation Notes

For biomechanics: specify sensor types, sampling rates, and units
For neuroperformance: specify cognitive domains, measurement modalities

Example Invocation Patterns

"Define a dataset schema for wearable sensor data in running studies (2024, field/lab, 100Hz sampling), including validation rules and harmonization notes."
"Propose a schema for cognitive task performance data, including metadata, versioning, and QA Validation Block."