phase-data¶

The data phase locates raw data, validates BIDS structure, and runs conversion scripts to get data ready for preprocessing.

Approach¶

Follow the three-step sequence in order: inventory → validate → convert — do not skip ahead
Confirm data location and expected modality before doing anything
For your own participants' data, check the ethics gate and ai_processing first (/data → Before you start)
Raw folders (raw_roots) are read-only — the rule and the BrainVision renaming trap are in /data step 3
BIDS compliance is the target; note deviations but do not block progress for minor issues
Summarize what was found, what passed, and what needs fixing before moving on

neuroflow:neuroflow-core — read first; defines the command lifecycle and .neuroflow/ write rules
neuroflow:bids — invoke whenever BIDS structure, validation, conversion, or sidecar metadata is relevant; provides full entity ordering, required files per modality, tool usage (bids-validator, pybids, MNE-BIDS), and conversion examples

Save data-inventory.md to .neuroflow/data/ listing datasets, paths, storage location type, and BIDS status
Conversion scripts and outputs go to output_path, not inside .neuroflow/
Log any structural data issues that affect analysis choices downstream in .neuroflow/reasoning/data.jsonl

scripts/erasure_sweep.py — finds every place a participant ID reached (project files, git history, Claude Code's local transcripts and caches, ~/.neuroflow/) and reports paths and counts; never deletes and never prints contents. Used by /ethics --erase; the person runs it in their own terminal: python <skill base dir>/scripts/erasure_sweep.py --id sub-07. Exit 0 = nothing found, 1 = occurrences found, 2 = usage error.

/neuroflow:data — runs this workflow as a slash command.