`/neuroflow:search`¶

Lightweight scoped search across project memory or the codebase.

/search lets you quickly find information without leaving your current workflow. It uses flow.md files as a fast index so it only reads what is relevant — no full directory scans.

Syntax¶

/search <tag>:<query>

Tags¶

Tag	Scope
`memory`	Search inside `.neuroflow/` only — plans, summaries, session logs, analysis notes, reasoning
`project`	Search the project codebase outside `.neuroflow/` — scripts, notebooks, results, config files

If no tag is supplied, the search defaults to memory and notifies you.

When to use it¶

You want to find where a decision was recorded without reading every file
You remember working on something (e.g. "ICA artifact rejection") and want to locate the relevant notes quickly
You want to find a script or notebook somewhere in the codebase
Another agent or command needs a quick answer about what the project memory contains

Examples¶

/neuroflow:search memory:ICA artifact rejection

🔍 Search: "ICA artifact rejection" [scope: memory]

Found in:
  • .neuroflow/data-preprocess/preprocessing-notes.md — ICA run and artifact components manually rejected
  • .neuroflow/reasoning/data-preprocess.json — decision: kept 58/64 components after visual inspection

Summary: ICA artifact rejection was completed during the data-preprocess phase. Notes are in
preprocessing-notes.md; the reasoning log records the decision and the number of components kept.

/neuroflow:search project:bandpass filter

🔍 Search: "bandpass filter" [scope: project]

Found in:
  • scripts/preprocessing/preprocess_eeg.py — apply_bandpass() function, 1–40 Hz
  • scripts/preprocessing/config.py — bandpass_low=1, bandpass_high=40 defined here

Summary: The bandpass filter is implemented in preprocess_eeg.py and its parameters are set in config.py.

How it works¶

Parses the tag and query from your invocation
Reads flow.md as a fast index — no full directory scans
For memory: navigates .neuroflow/ phase subfolders guided by the index
For project: uses output_path entries from flow.md to locate relevant directories, then searches there
Returns a compact summary: matched files with one-line reasons + a 2–4 sentence synthesis

This command is read-only — it never writes to .neuroflow/ or any project file.

Files read and written¶

Direction	Files
Reads	`.neuroflow/project_config.md`, `.neuroflow/flow.md` (and phase sub-`flow.md` files as needed)
Writes	(nothing — read-only command)

/neuroflow — full project status
/phase — show current phase and switch
/sentinel — full audit of .neuroflow/ for consistency issues

/neuroflow:search¶