/neuroflow:doctor¶
A health check of the setup around a neuroflow project: tools, backups, storage location, contracts — and whether the neuroflow mod is live.
What it checks¶
| Check | Why it matters |
|---|---|
| Python 3.10+, git, Claude Code version | neuroflow's scripts and the mod need them |
| git repository, remote, unpushed commits, last commit age | project memory that exists on one disk only is one disk failure from lost |
uncommitted changes under .neuroflow/ |
the same, for today's work |
| network share or cloud-synced folder | sync and shares corrupt or duplicate files written mid-sync |
project_config.md frontmatter and nf_schema |
the mod and the scripts read the current contract |
the neuroflow version the project is on (plugin_version) against the installed one |
after an update, the project's memory may still be in an older format — /migrate brings the project, your flowie and the team hive up to date |
.gitignore and .gitattributes |
local-only files stay out of git; append-only logs merge cleanly |
your flowie, if set up: unpushed commits, failures in ~/.neuroflow/flowie-sync.log (count and first timestamp) |
changes that never reached your private repo exist on this machine only — /flowie --sync resolves both; nothing is reported without flowie |
| your flowie, if set up: uncommitted changes (a count and up to three paths, never file contents) | they exist on this machine only, and changes to tracked files stop every pull, so your flowie stops updating — /flowie --sync commits and pushes them |
Is the mod live?¶
When the neuroflow mod is running, it answers /neuroflow:doctor itself, instantly: the first
lines name its settings and any feature that could not run. If the answer comes from the model instead, the mod is not
live in that session — the report says so.
The same checks run either way, from one script: skills/neuroflow-core/scripts/doctor.py (--json for machine
output; exit 0 all good, 1 warnings, 2 could not run).