/neuroflow:dashboard¶
The project at a glance: phase map, upcoming deadlines, integrity state, open tasks at every level and the running autoresearch loop.
With the neuroflow mod¶
The dashboard opens as a pane drawn by code — no model turn, no tokens — docked beside the conversation in a wide terminal, or inline above the prompt otherwise. It opens holding the keyboard, so its keys work at once, and it stays current as files change.
| Key | Tab |
|---|---|
1 |
Now — everything the band above the prompt can show, whatever its band setting, each with its action: a drafted decision to keep or drop, a meeting to prepare or close, a flowie sync, wiki cards to review, the loop being driven, the next phase; under them the phase, task and loop lines (beside them in a wide pane). An action is Enter on its row, never a letter, and asks first in Claude Code's question dialog with the safe answer on top (the workbench's apply only puts its command in the prompt); j and k move between the rows, starting on the tab's own name, and an Enter that comes right after another key does nothing — so a prompt typed into the pane by mistake does not migrate, keep or sync anything without your yes |
2 |
Phase — the phases as a flow line (◉ current, ● visited, → recommended next) with the current and the next step; g opens the phase picker |
3 |
Deadlines — a ruler of the coming weeks with each date's mark on its day (⚠ within three days, ▸ within two weeks, · later), then the dated rows of .neuroflow/timeline.md and the ethics expiry, with days left and the phase each gates |
4 |
Integrity — ethics approval, frozen preregistration, read-only raw-data folders, config problems; Enter on freeze prereg… freezes the preregistration after your explicit yes, v re-checks its hashes, Enter on unfreeze… unfreezes it with a reason (Tab reaches both: setting a person's marker never hangs on a letter) |
5 |
Tasks — open tasks per level on one line (project 3 · flowie 12 (2 Oddball EEG) · example-lab-hive 4: the project, your flowie and each hive, with this project's share where that level links it, then the project's other repositories from the local cache of their boards), then the first open tasks across levels, each tagged with its level — this project's first, overdue first; b opens the boards |
6 |
Loop — the running autoresearch loop: iteration, best snapshot, a quality sparkline and its open questions |
On every tab h and l go to the previous and the next tab, r reads the files again, q closes the dashboard and
Esc gives the keys back to the prompt (the pane stays). /neuroflow:dashboard loop opens it on a tab, and running
/neuroflow:dashboard again gives an open dashboard the keys again. The band above the prompt opens it with d: on the
now tab, or on the loop tab while a loop is driven; the band keeps the keys, so run the command to give them to it.
When the project is on an older neuroflow than the one installed, an Update line sits under the title, with a migrate
action that runs /neuroflow:migrate on Enter and your yes — or, while a wellbeing check-in still waits
to sync to your flowie, puts it in the prompt: press Esc and then Enter, and the sync runs first, so the migration finds
your flowie clean. All keys, the chord that reaches the band, and working over SSH:
The neuroflow mod → Keyboard and SSH.
Every symbol comes with a word, so the dashboard reads correctly without colour and with a screen reader. The pane lays itself out for its width — the now tab side by side where the pane gets 90 cells or more (inline in a terminal of about 95 columns and up), stacked in a dock at its own width — and drops its blank rows on a terminal under 30 rows, such as 24 rows over SSH.
Without the mod¶
The same overview is printed once as a compact text block, read from the same files. Nothing is written. When the
project is on an older neuroflow version than the one installed, an Update line names /neuroflow:migrate;
with the mod, the band above the prompt and the dashboard pane say the same.
--web — your claude.ai workbench¶
/neuroflow:dashboard --web publishes and refreshes your workbench: one private claude.ai
page with a tab for your flowie, each hive you joined and each project you export it from — a board with drag and drop
and keyboard moves, the project list, the wiki and the level's panel. Git stays the only store; a click on the page is
a request that /neuroflow:tasks --apply-web checks and applies.
| Mode | What it does |
|---|---|
--web |
The first time: explains what would leave your machine, asks your consent, a name for this machine and which levels to mirror, then — after a one-line preview and your yes — publishes the page and uploads. After that: refreshes the levels, upgrades the page when the installed neuroflow ships a newer one, cleans up finished requests, deletes a level whose team no longer allows mirroring it (the preview names it), and copies your page settings into your flowie |
--web --status |
Which account and page, the page's version against the plugin's, each level's consent and the team's opt-in, whether ID numbers and participant names are checked (pii_config), how many changes wait and since when — nothing is uploaded |
--web --levels |
Choose again which levels are mirrored; one you take off is purged from the page |
Every upload asks first, every time: nothing is sent on a timer or at the end of another command. A hive or project is
mirrored only when its team set workbench_mirror (in hive.md or project_config.md); without it, it is left out
and the command says why. With the mod active, --web runs the full command too.
Files read¶
| Direction | Files |
|---|---|
| Reads | .neuroflow/project_config.md, .neuroflow/timeline.md, .neuroflow/ethics/status.md, .neuroflow/preregistration/status.md, .neuroflow/tasks/, ~/.neuroflow/flowie/tasks/ and each ~/.neuroflow/hives/{org-repo}/tasks/ (the local copies, never pulled), ~/.neuroflow/local-projects.json and ~/.neuroflow/flowie/projects/projects.json (which flowie project this is), each hive's projects/projects.json (which lab project it is there), ~/.neuroflow/user.yaml (hive order), .neuroflow/{phase}/autoresearch-loops.md and the loops' results.md / report.md; with --web also the levels it mirrors, ~/.neuroflow/workbench/ and the workbench record (~/.neuroflow/flowie/workbench.json) |
| Writes | nothing, unless you freeze or unfreeze the preregistration from the integrity tab: then freeze.py writes .neuroflow/preregistration/status.md, the banners and deviations.md, and the mod adds a session line and a reasoning entry; keeping a drafted decision on the now tab adds its entry to .neuroflow/reasoning/{phase}.jsonl. --web writes your consent, levels, machine name and pii_config path to ~/.neuroflow/user.yaml, ~/.neuroflow/workbench/pending.json, the workbench record, and a session line — and uploads, after your yes, to your own claude.ai workbench |
Related¶
/phase— the phase map and phase switching (a picker with the mod)/tasks— work the task boards (with the mod, every level in one pane)- The neuroflow mod — what the mod adds and how to turn it on or off