v1.2.1

What's new

• README template expanded from 3 sections to the full 9-section
  structure (overview, experimental design, dataset contents,
  methods breakdown, known issues, references, contact), matching
  official BIDS README guidance instead of a shortened version.
• Added HEDVersion to the dataset_description.json template —
  was documented as RECOMMENDED in bids_reference.md but missing
  from the actual template.

Full Changelog: v1.2.0...v1.2.1

v1.2.0

What's new

• SKILL.md and references compressed to caveman style, cutting the
  skill's load cost by ~8%. Verified against a 21-question fidelity
  quiz before merging — no loss found.
• Frontmatter description trimmed (125 -> 75 words). It stays resident
  in context every session, unlike the skill body.

Full Changelog: v1.1.0...v1.2.0

v1.1.0

What's new

• The skill now reads only the material relevant to your conversion
  scenario, instead of pulling in unrelated reference text.
• Two outdated instructions were fixed: the skill no longer treats "the
  validator passed" as proof the conversion is correct, and no longer
  assumes a metadata field is blank when it sometimes isn't.
• Sidecar files are now checked for misspelled fields, not just wrong
  values, so mistakes that used to pass silently get caught.
• Signal verification checks amplitude separately from shape now, catching
  unit mix-ups (like microvolt vs. volt) that used to look "verified."
• Missing file templates added for cases the spec allows but the skill
  didn't cover yet.
• Dataset citation and funding info can now be recorded during conversion.

Fixes

• A flat or disconnected channel no longer aborts an otherwise correct
  conversion.
• Datasets using their own labels (e.g. m/f) no longer get flagged as if
  those were BIDS field names.
• Recordings exported from Windows no longer get counted and converted
  twice.

Full Changelog: v1.0.0...v1.1.0

v1.0.0: First Release

Overview

Specialized skill for AI agents to automate organization of EEG datasets into Brain Imaging Data Structure (BIDS) compliant format.

Key Features

• Automated BIDS Conversion: AI-driven conversion of raw recording files to standardized BIDS hierarchy.
• Metadata Management: Tools for agents to generate and patch JSON sidecar files for precise data description.
• Dataset Validation: Built-in validation ensuring AI-organized data meets EEG-BIDS specifications.
• Dataset Inspection: Utility to audit and verify dataset structure and content.

Components

• convert_recording.py: Primary conversion engine.
• write_bids_metadata.py: BIDS metadata generator.
• patch_sidecar.py: Sidecar modifier for specific parameter updates.
• validate_bids.py: BIDS compliance checker.
• inspect_dataset.py: Dataset summary and structure auditor.

Usage

1. Download this repository.
2. Place in your local skills folder.
3. Invoke via: /eeg-bids-converter