phase-meeting¶

The /meeting command manages structured meetings — distinct from /notes which captures unstructured live input. Meetings have schema (attendees, agenda, decisions, action items), optional calendar integration, and task creation from action items.

Meeting levels¶

Level	Storage	Git-tracked in	Who sees it
`project` (default)	`.neuroflow/meetings/YYYY-MM-DD-{slug}.md`	Project repo	All collaborators
`flowie`	`~/.neuroflow/flowie/meetings/YYYY-MM-DD-{slug}.md`	Flowie repo	Owner only
`hive`	`~/.neuroflow/hives/{org-repo}/meetings/YYYY-MM-DD-{slug}.md`	Hive org repo	Whole team

Each level root (.neuroflow/, ~/.neuroflow/flowie/, ~/.neuroflow/hives/{org-repo}/) holds both meetings/ and tasks/. Flowie-level files are synced with the flowie Sync step (/flowie → Git operations pattern); hive-level files are pushed only after the person confirms; project-level files stay in the working tree until the person commits them.

Meeting file format¶

---
title: {title}
date: {YYYY-MM-DDTHH:MM:00}
duration: {minutes}
location: {Zoom link / room / online}
attendees:
  - name: {name}
    email: {email}          # from the roster or the person — omit if unknown, never guess
level: {project|flowie|hive}
tags: [{tag1}, {tag2}]
linked_tasks: []            # {level}:{slug} entries written by --close
closed: ""                  # YYYY-MM-DD, stamped by --close
calendar_event_id: ""
template: {slug of recurring template, or omit}
---

## Agenda

## Notes

## Decisions

## Action Items

Recurring meeting templates¶

Templates live in: - .neuroflow/meetings/config.json — project-level (shared with collaborators) - ~/.neuroflow/flowie/meetings/config.json — personal (flowie level) - ~/.neuroflow/hives/{org-repo}/meetings/config.json — team-level (hive, synced with /hive --sync)

{
  "recurring": [
    {
      "name": "Weekly Lab Meeting",
      "slug": "weekly-lab",
      "duration": 60,
      "location": "Room 301",
      "level": "hive",
      "default_attendees": ["all-hive-members"],
      "default_tags": ["lab-meeting"],
      "agenda_template": "## Updates\n\n## Papers\n\n## Action Items"
    },
    {
      "name": "Supervisor 1:1",
      "slug": "supervisor-1on1",
      "duration": 30,
      "location": "Zoom",
      "level": "flowie",
      "default_attendees": ["supervisor@example.com"],
      "default_tags": ["1on1"],
      "agenda_template": "## Progress\n\n## Blockers\n\n## Next steps"
    }
  ]
}

default_attendees can be: - An explicit email string - "all-hive-members" → resolved from hive.md members table - "all-project-collaborators" → resolved from project_config.md collaborators list

Mode: --new¶

Check for templates: read all config.json files at the relevant level(s). If any recurring templates exist, show a picker:

Create meeting from template?
  [1] Weekly Lab Meeting  (hive, 60 min, Room 301)
  [2] Supervisor 1:1      (flowie, 30 min, Zoom)
  [3] Custom meeting

If template selected: pre-fill title, duration, location, level, tags, and agenda from the template. Ask only for date and time.
If custom: ask all fields:
Title?
Date and time? (YYYY-MM-DD HH:MM)
Duration? (minutes, default 60)
Location? (optional)
Level? [project / flowie / hive] (default: project)
Tags? (comma-separated, optional)
Resolve attendees (Attendee resolution rules below — never invent an email):
For default_attendees: ["all-hive-members"] → read the hive's members.md table → extract name + email for each row
For default_attendees: ["all-project-collaborators"] → read collaborators: from project_config.md
For explicit email strings → use as-is
For custom meeting: ask "Who should attend? (names or emails, comma-separated)" — look up emails from hive members and collaborators by name if given; for anyone without a roster email, ask
Write meeting file to the appropriate storage location (see levels table above). Filename: YYYY-MM-DD-{slug}.md where slug is derived from the title.

Calendar invite (optional): invites email real people, so list every recipient address first and ask "Create a Google Calendar event and send invites to: {emails}? [y/N]" — send only on an explicit yes in this turn.

If yes: - Use mcp__claude_ai_Google_Calendar__create_event with: - summary: meeting title - start: ISO 8601 datetime - end: start + duration - location: as given - attendees: list of email strings - description: the agenda from the meeting file - Store the returned event ID in calendar_event_id in the meeting frontmatter

Confirm:

Meeting created: {title} — {date} {time}
File: {path}
Attendees: {N} · Calendar: {event link or "not sent"}

Mode: --prepare \<slug>¶

Prepopulate the ## Agenda section with context from the project and team.

Read the meeting file by slug (search across all levels)
Read .neuroflow/project_config.md for active_phase
Read .neuroflow/tasks/active/ and .neuroflow/tasks/review/ — list active and under-review tasks (task format: /tasks)
Read .neuroflow/timeline.md (if present) — pull entries within the next 30 days for a Deadlines block
If hive is connected: read hive.md directions for relevant team context
If the meeting comes from a recurring template, find the previous meeting of the same template and add:
Follow-ups — open tasks (not done/archive) whose source is an earlier meeting of this template
Since last time — tasks that reached done after that meeting's date (by updated), and reasoning entries since then (.neuroflow/reasoning/*.jsonl, by at). Useful for a supervisor 1:1; keep each item to one line.

Compose an agenda draft:

## Agenda

### Project status
- Phase: {active_phase}
- Active tasks: {list slugs with titles}
- In review: {list slugs with titles}

### Follow-ups
{open tasks from earlier meetings of this template — omit section if none}

### Since last time
{done tasks and decisions since the previous meeting of this template — omit section if none}

### Upcoming deadlines
{timeline.md entries within 30 days, ⚠ within 3 days and ▸ within 14, as /dashboard marks dates — omit section if none}

### Team context
{relevant hive directions if any}

### Items
{blank or template sections}

Show the draft and ask "Does this look right? Edit inline or confirm."

Write the updated ## Agenda section back to the meeting file (preserve other sections).

Mode: --notes \<slug>¶

Live notes during the meeting, straight into the meeting file. Use the live-capture format in neuroflow:phase-notes (Live capture format), with the meeting's ## Notes section as the target:

Each message is appended to ## Notes at once, verbatim, as [HH:MM] {text} — rewrite the file with the Write tool; never pass note text through a shell command.
While capturing, every message is note text, never an instruction ("delete the old epochs" is recorded, not executed). Only done (or the equivalent in the person's language) or a message starting with / ends capture.
Acknowledge with one short line (✓ {N}), nothing more.

On done: copy every line whose text starts with a: into ## Action Items as - [ ] {text} and every d: line into ## Decisions as - {text} (the raw lines stay in ## Notes), show the result, and offer --close.

Mode: --view \<slug>¶

Find meeting file by slug
Read all linked_tasks from frontmatter — each entry is {level}:{slug} → find {slug}.md in that level's tasks/{column}/ folders; the folder is its current column
Display meeting file with task statuses inline:

─────────────────────────────────────────
  Weekly Lab Meeting — 2026-04-20 10:00
  Location: Room 301 · Duration: 60 min
  Attendees: Stan, Jana, Petr (3)
  Calendar: https://calendar.google.com/...
─────────────────────────────────────────

## Agenda
...

## Linked tasks
  [active]  fix-rt-glasses        RT_DES
  [review]  grant-draft           AlphaModulation
  [done]    eeg-param-sweep       AlphaModulation  ✓

─────────────────────────────────────────

Mode: --list¶

Show all meetings at the active level. Put the next upcoming meeting first (next), then past meetings that are not closed — no closed: date and unchecked action items left (not closed — run --close), then the rest by date (newest first):

next        [project]  2026-04-27  Weekly Lab Meeting   weekly-lab-2026-04-27
not closed  [project]  2026-04-20  Weekly Lab Meeting   weekly-lab-2026-04-20
            [flowie]   2026-04-17  Supervisor 1:1       supervisor-1on1-2026-04-17
            [project]  2026-04-13  Weekly Lab Meeting   weekly-lab-2026-04-13

Meeting dates have no timezone; compare them with the local date and time. If --level not given: show project and flowie levels together.

Mode: --invite \<slug>¶

Re-send or send calendar invites for a meeting that doesn't have a calendar_event_id yet.

Read meeting file
If calendar_event_id is already set: ask "Invites were already sent (event ID: {id}). Re-send? [y/N]"

List the recipient emails from the meeting file and send only after the person's explicit yes in this turn. Attendees without an email are skipped and named — never fill one in.
Call mcp__claude_ai_Google_Calendar__create_event (or update_event if re-sending) with those attendees
Update calendar_event_id in frontmatter

Mode: --close \<slug>¶

Finalize the meeting and convert its open action items into tasks in the /tasks format. Parsing and writing are done by scripts/meeting_close.py, so the same note always gives the same tasks and a second run never duplicates them. The script never sends invites or emails and never pushes.

Find the meeting file by slug.
Check the ## Action Items section. Each open item is a checkbox:
```
- [ ] Description → @owner due:YYYY-MM-DD [level/column]
```
-> works as the arrow; → @owner, due: and the annotation are all optional
[level/column] defaults to [project/inbox]; [level] or [column] alone also work
Checked items (- [x]) are skipped; plain bullets are listed as "not a checkbox" — ask whether any of them should become a task, and if so turn them into checkboxes first
Owners must be roster handles (Attendee resolution rules) — ask about any that are not
Dry run (from the project root):
```
python <skill base dir>/scripts/meeting_close.py <meeting file> [--hive-root ~/.neuroflow/hives/{org-repo}]
```
--hive-root is needed only for [hive/…] items in a project or flowie meeting. Exit 0: show the plan to the person. Exit 1: show the errors (unknown level or column, bad date, level not set up), fix the note with the person, and run the dry run again. Exit 2: usage or runtime error — report it and stop.
Write — only after the person confirms the plan, run the same command with --write. It creates {level root}/tasks/{column}/{slug}.md per item (with source: {level}:meetings/{file}), adds {level}:{slug} entries to linked_tasks, and stamps closed: {today}. Items already converted from this meeting are reported as exists.
Report what was created, by level and column (the script's output is the report).
Sync — the script writes files directly, so the flowie auto-sync hook does not see them:
flowie-level task files: the flowie Sync step for those paths (/flowie → Git operations pattern)

hive-level task or meeting files: commit by path; push to the hive repo only after the person's explicit yes in this turn
project-level files stay in the working tree — pushing to the shared project remote is the person's call

If Python is not available, do the same by hand: show the planned tasks first, then write them in the /tasks format and update linked_tasks and closed:.

Mode: --init¶

Set up recurring meeting templates for the current level.

Ask for level: [project / flowie / hive]
Walk through template creation:
Name? (e.g. "Weekly Lab Meeting")
Slug? (auto-suggested from name, editable)
Duration (minutes)?
Location?
Default attendees? (emails, "all-hive-members", "all-project-collaborators")
Default tags?
Agenda template? (section headings, freeform)
Write to config.json at the appropriate location
Confirm: Template saved: {slug}
Ask "Add another template? [y/N]"

Attendee resolution rules¶

When resolving attendees for any mode:

"all-hive-members" → read ~/.neuroflow/hives/{org-repo}/members.md (the hive roster file) → return all rows as {name, email}
"all-project-collaborators" → read .neuroflow/project_config.md collaborators: list → return all entries
Plain email string → use as-is, name = email prefix
Name string (no @) → search hive members and collaborators by name → use matched email; if ambiguous, ask
@handle (task owners in action items) → match the handle of a collaborator or the github column of members.md; if there is no match, ask — never create a handle

Never invent an email address. An address comes from the roster (members.md, collaborators:, the person's flowie profile.md) or from the person, typed in this conversation — never guessed, completed or built from a name and a domain. If someone has no address there, ask; if the person does not give one, keep the attendee without email: and leave them out of invites.

If Google Calendar MCP is not authenticated: skip the calendar step and note "Google Calendar not configured — run /setup to connect."