tmr-inbox

name	tmr-inbox
description	Process Granola meeting notes from inbox/ — route to correct folder, rename with consistent naming, scaffold missing person profiles, extract tasks to my-tasks/tasks.md

tmr-inbox

Processes all .md files in inbox/ using a two-pass batch approach.

Usage

• /tmr-inbox — run the full inbox triage
• /tmr-inbox setup — one-time Dataview install + task view configuration (run once before first use)

When invoked, check if the user passed the argument setup. If so, follow the SETUP section. Otherwise follow the TRIAGE section.

Before either command, BOOTSTRAP is always run first to resolve MANAGER_EMAIL and INTERNAL_DOMAINS.

Vault owner: resolved dynamically in BOOTSTRAP — see Step BT2.

---

BOOTSTRAP

Step BT1: Locate vault root

git rev-parse --show-toplevel

Store the result as VAULT_ROOT.

Step BT2: Resolve vault owner email

List .md files in my-career/ non-recursively. Identify the one whose name contains @. Store its filename stem (without .md) as MANAGER_EMAIL.

If none is found, exit with: tmr-inbox: could not resolve vault owner email from my-career/ — run \tmr init` to set up your vault.`

Step BT3: Resolve internal domains

Read config/organization.yaml if it exists. Extract the internal_domains list as INTERNAL_DOMAINS.

If the file does not exist or the internal_domains field is absent, set INTERNAL_DOMAINS to a single-element list containing the domain portion of MANAGER_EMAIL (the part after @).

---

SETUP

Step S1: Install Dataview Plugin

Run from the vault root:

mkdir -p ".obsidian/plugins/dataview"
curl -sL "https://github.com/blacksmithgu/obsidian-dataview/releases/latest/download/main.js" \
  -o ".obsidian/plugins/dataview/main.js"
curl -sL "https://github.com/blacksmithgu/obsidian-dataview/releases/latest/download/manifest.json" \
  -o ".obsidian/plugins/dataview/manifest.json"
curl -sL "https://github.com/blacksmithgu/obsidian-dataview/releases/latest/download/styles.css" \
  -o ".obsidian/plugins/dataview/styles.css"

Step S2: Enable Dataview in Obsidian

Read .obsidian/community-plugins.json. It is a JSON array of enabled plugin IDs. If "dataview" is not in the array, add it and write the file back.

Example before: ["obsidian-git", "calendar"] Example after: ["obsidian-git", "calendar", "dataview"]

Step S3: Create tasks.md if missing

Check if my-tasks/tasks.md exists. If not, write it:

---
objective: Canonical task list — source of truth for all action items
---

# Tasks

Step S4: Rewrite my-tasks view files

Write these four files exactly (overwrite existing content):

my-tasks/today.md:

---
objective: Tasks that need to be accomplished today
---

# Today

```dataview
TASK
FROM "my-tasks/tasks.md"
WHERE !completed AND contains(text, "#today")
SORT file.mday DESC

**my-tasks/this-week.md:**
```markdown
---
objective: Tasks that need to be accomplished this week
---

# This Week

```dataview
TASK
FROM "my-tasks/tasks.md"
WHERE !completed AND contains(text, "#this-week")
SORT file.mday DESC

**my-tasks/this-month.md:**
```markdown
---
objective: Tasks that need to be accomplished this month
---

# This Month

```dataview
TASK
FROM "my-tasks/tasks.md"
WHERE !completed AND contains(text, "#this-month")
SORT file.mday DESC

**my-tasks/this-quarter.md:**
```markdown
---
objective: Tasks that need to be accomplished this quarter
---

# This Quarter

```dataview
TASK
FROM "my-tasks/tasks.md"
WHERE !completed AND contains(text, "#this-quarter")
SORT file.mday DESC

### Step S5: Print completion message

Print exactly:

tmr-inbox setup complete.

Installed: .obsidian/plugins/dataview/ — Dataview plugin files my-tasks/tasks.md — canonical task list my-tasks/today.md — live Dataview view my-tasks/this-week.md — live Dataview view my-tasks/this-month.md — live Dataview view my-tasks/this-quarter.md — live Dataview view

ACTION REQUIRED: Restart Obsidian to activate the Dataview plugin. After restart, open any file in my-tasks/ to see live task queries.

Run /tmr-inbox to process your inbox.

---

## TRIAGE

### Pass 1: Scan and Classify (partial reads)

#### T1. Check for empty inbox

Use Glob to find all `.md` files in `inbox/`. If none found, print:

Inbox is empty — nothing to process.

Stop.

#### T2. Classify all files in one batch

For each inbox file, read only the frontmatter and first 60 lines (do NOT read the full transcript in this pass — conserve tokens for execution).

Extract from each file:
- `attendees` array from YAML frontmatter
- `title` string from YAML frontmatter
- `created` timestamp from YAML frontmatter (extract YYYY-MM-DD prefix as the meeting date)
- First heading and first 5-10 bullet points from content

**Exclude MANAGER_EMAIL from attendee counts.**

Classify each file using these rules in order:

**Rule 1 — 1:1 check:**
After removing MANAGER_EMAIL from attendees, is exactly 1 person left?
- Yes → determine their folder:
  - Check `my-teams/members/[email]/` exists → type: `1on1-team`
  - Else check `my-leadership/[email]/` exists → type: `1on1-leadership`
  - Else check `my-company/contractors/[email]/` exists → type: `1on1-contractor`
  - Else → type: `1on1-member`
- Destination: `my-teams/members/[email]/1on1s/`, `my-leadership/[email]/1on1s/`, `my-company/contractors/[email]/1on1s/`, or `my-company/members/[email]/1on1s/`

**Rule 2 — Project meeting check:**
2+ other attendees remain. Does the title or content strongly suggest an existing project?
- Use Glob to list all folder names under `my-company/projects/`
- Normalize: lowercase the title, strip accents, replace spaces with hyphens
- If the normalized title contains a project folder name (or vice versa) → type: `project`, destination: `my-company/projects/[matched-project]/meetings/`
- If no folder match but content clearly names a recognizable project → type: `project`, use the inferred project name

**Rule 3 — Company/general meeting:**
3+ other attendees, no clear project match → type: `company`, destination: `my-company/meetings/`

Note: a meeting with exactly 2 other attendees and no project match does NOT satisfy Rule 3. It falls to Rule 4 (ambiguous) and will be presented to the user for routing.

**Rule 4 — Ambiguous:**
Cannot determine type with confidence (e.g., 2 attendees but both are unknown, title is vague, project match is weak) → type: `ambiguous`

Build two lists:
- `confident`: [{file, type, destination, person_email or project_name, date}]
- `ambiguous`: [{file, title, attendees, reason}]

#### T3. Resolve ambiguity (if any)

If the `ambiguous` list is non-empty, process each one interactively before executing anything.

For each ambiguous file, display:

--- Ambiguous file --- File: [original filename] Title: [title from frontmatter] Attendees: [list, MANAGER_EMAIL excluded] Reason: [why it was flagged — e.g., "2 attendees but neither found in vault", "title unclear"]

How should this be routed? a) 1:1 with [first attendee email] → my-company/members/[email]/1on1s/ b) Project meeting — enter project name: ___ c) Company/general meeting → my-company/meetings/ d) Skip this file e) Contractor 1:1 with [first attendee email] → my-company/contractors/[email]/1on1s/

Wait for the user's response. Apply:
- `a` → add to confident as type `1on1-member` with that email
- `b` → add to confident as type `project` with the entered project name
- `c` → add to confident as type `company`
- `d` → remove from all queues (file stays in inbox); add to `skipped` list with reason 'user deferred'
- `e` → add to confident as type `1on1-contractor` with that email

After all ambiguous files are resolved, proceed to Pass 2.

### Pass 2: Execute (full reads)

For each file in the `confident` list, perform steps 2a through 2h in sequence.

#### Step 2a: Resolve attendee emails

Build a `resolved_attendees` map for all entries in the `attendees` frontmatter array (excluding MANAGER_EMAIL).

For each attendee entry:

**If it already contains `@`** (it is an email address) → add to map as-is, `inferred: false`.

**If it is a full name only** → infer the email:

1. **Determine the dominant domain**: collect all email-format attendees in this same meeting. Count unique domains. Use the most frequent domain. If there are no email-format attendees, default to `@<first domain in INTERNAL_DOMAINS>`.

2. **Check for an existing profile that matches the name**: scan folder names in `my-teams/members/`, `my-leadership/`, `my-company/contractors/`, and `my-company/members/`. For each folder whose name looks like `firstname.lastname@domain`, check if the `firstname` and `lastname` match the person's first given name and last surname (strip accents, lowercase, compare). Also check the `name` field in the profile's frontmatter if present. If a matching folder is found → use that email, `inferred: false` (it's a confirmed match).

3. **If no existing profile matches** → construct the inferred email:
   - Strip Portuguese accents from the full name: ã→a, á→a, à→a, â→a, é→e, ê→e, í→i, ó→o, ô→o, õ→o, ú→u, ç→c, ñ→n
   - Take the **first word** (first given name) and the **last word** (last surname)
   - Lowercase both, combine as `firstname.lastname@dominant-domain`
   - Add to map with `inferred: true`

Examples:
- "John Gran Doe" → `[e-mail]` (inferred)
- "Mark Schward" → `[e-mail]` (inferred)
- "James Dough" → `[e-mail]` (inferred)
- "[e-mail]" → kept as-is (inferred: false)

Use this `resolved_attendees` map for all subsequent steps.

#### Step 2b: Scaffold missing person profiles

**For 1:1 types** (`1on1-team`, `1on1-leadership`, `1on1-contractor`, `1on1-member`):

Check if the person's folder exists:

**If type is `1on1-team`:** check `my-teams/members/[email]/`
**If type is `1on1-leadership`:** check `my-leadership/[email]/`
**If type is `1on1-contractor`:** check `my-company/contractors/[email]/`
**If type is `1on1-member`:** check `my-company/members/[email]/`

If the folder does NOT exist, create it:

For `my-teams/members/[email]/`:
```bash
mkdir -p "my-teams/members/[email]/1on1s"
mkdir -p "my-teams/members/[email]/assessments"
mkdir -p "my-teams/members/[email]/feedbacks"
mkdir -p "my-teams/members/[email]/performance-reviews"

For my-leadership/[email]/:

mkdir -p "my-leadership/[email]/1on1s"

For my-company/contractors/[email]/:

mkdir -p "my-company/contractors/[email]/1on1s"

For my-company/members/[email]/:

mkdir -p "my-company/members/[email]/1on1s"

For project type:

First, check if my-company/projects/[project-name]/ exists. If it does NOT, create the project scaffold:

mkdir -p "my-company/projects/[project-name]/meetings"

Then create my-company/projects/[project-name]/[project-name]-project.md:

---
name: [project-name]
status: active
start_date: <YYYY-MM-DD>
owner: unknown
---

# [project-name]

First meeting: [[new-filename-without-extension]]

Where start_date = the meeting date extracted from the created field (same date used in step 2c). owner: unknown is an explicit signal that the skill did not have manager profile context at creation time — tmr-myself-config will overwrite this when run. Where new-filename-without-extension is the filename that will be computed in step 2c (without .md). Track any newly created projects in a scaffolded_projects list for the final summary.

For project and company types:

For each resolved attendee (from the map built in step 2a), excluding Marlon:

1. Check if a folder already exists in any of: my-teams/members/[email]/, my-leadership/[email]/, my-company/members/[email]/
2. If no folder found → create:

mkdir -p "my-company/members/[email]/1on1s"

Step 2c: Determine new filename

Apply the naming pattern for the file's type.

For 1:1 types (1on1-team, 1on1-leadership, 1on1-member):

YYYY-MM-DD-1on1-[email].md

• YYYY-MM-DD = meeting date extracted from the created frontmatter field
• [email] = the other person's email (lowercase, as-is)
• Example: [e-mail]

For project type:

YYYY-MM-DD-[kebab-topic]-[project-name]-project.md

• [kebab-topic] = title lowercased, Portuguese accents stripped (ã→a, ç→c, é→e, etc.), spaces→hyphens, non-alphanumeric/hyphen chars removed; take the first 5 words from the original title before normalization, then apply the transformations
• [project-name] = matched or user-provided project folder name (already kebab-case)
• Literal suffix: -project
• Example: 2026-04-10-communication-plan-ai-enablement-project.md

For company type:

YYYY-MM-DD-[kebab-topic].md

• [kebab-topic] = same normalization as above (take the first 5 words from the original title before normalization, then apply the transformations)
• Example: 2026-04-10-all-hands-q2-review.md

Collision check: if the computed destination path already exists, append -2, then -3, etc. until the path is unique.

After computing the new filename, for each resolved attendee: if the profile file [their-folder]/[email].md does NOT exist yet, create it now.

For attendees whose email was not inferred (confirmed email):

---
name: [Full Name or email if no display name available]
email: [email]
relationship: <derived from routing context>
---

First seen: [[new-filename-without-extension]]

Derive relationship from the routing context:

• 1on1-team → direct-report
• 1on1-leadership → leadership
• 1on1-contractor → contractor
• 1on1-member, project, or company attendee with domain in INTERNAL_DOMAINS → company
• 1on1-member, project, or company attendee with domain not in INTERNAL_DOMAINS → unknown

(INTERNAL_DOMAINS is resolved in BOOTSTRAP Step BT3.)

For attendees whose email was inferred from their name:

---
name: [Full Name as it appears in attendees]
email: [inferred-email]
email_inferred: true
relationship: unknown
---

First seen: [[new-filename-without-extension]]

Inferred-email profiles are always relationship: unknown — their identity and relationship to the manager are unverified.

Where new-filename-without-extension is the filename just computed above, without .md.

Step 2d: Read full file content

Read the complete inbox file. You now have the full transcript and action signals.

Step 2e: Enrich frontmatter and inject backlinks

Part 1 — Enrich the YAML frontmatter block.

Before writing the file to its destination, add the following fields into the existing YAML frontmatter (between the opening --- and closing ---). Append them at the end of the existing frontmatter fields — do not remove or alter any existing fields (title, attendees, created, etc.):

tmr_type: <type>          # one of: 1on1 | project | company
date: <YYYY-MM-DD>        # extracted from the `created` field (date portion only)
participants:             # resolved emails only, Marlon excluded
  - <email1>
  - <email2>
project: <slug>           # only present for tmr_type: project; omit for all others

Use tmr_type (not type) to avoid collision with any Obsidian or Granola reserved field.

Example result for a project meeting:

---
title: AI Enablement — Communication Plan
attendees:
  - [e-mail]
  - John Gran Doe
created: 2026-04-10T14:00:00
tmr_type: project
date: 2026-04-10
participants:
  - [e-mail]
  - [e-mail]
project: ai-enablement
---

Example result for a 1:1:

---
title: 1:1 with Alice
attendees:
  - [e-mail]
created: 2026-04-10T09:00:00
tmr_type: 1on1
date: 2026-04-10
participants:
  - [e-mail]
---

Part 2 — Inject wiki-link backlinks.

After the closing ---, immediately insert the backlink block (no blank line between --- and People:):

For all types:

People: [[[e-mail]]] [[[e-mail]]]

Include all attendees except Marlon. Use the resolved emails from the resolved_attendees map built in step 2a (not the raw frontmatter values — name-only entries must appear as their resolved [[email]] links).

For project type, add a second line:

Project: [[project-name]]

Add one blank line after the backlink block (before the first heading). Do not alter any existing content below it.

Resulting structure:

---
[existing Granola frontmatter]
tmr_type: 1on1
date: 2026-04-10
participants:
  - [e-mail]
---
People: [[[e-mail]]]

### [existing heading]
...

Or with project:

---
[existing Granola frontmatter]
tmr_type: project
date: 2026-04-10
participants:
  - [e-mail]
  - [e-mail]
project: ai-enablement
---
People: [[[e-mail]]] [[[e-mail]]]
Project: [[ai-enablement]]

### [existing heading]
...

Step 2f: Extract action items and append to tasks.md

Scan the full content for action items. Identify text that matches any of:

• Lines already formatted as - [ ] ...
• Lines containing: "follow up", "follow-up", "send", "schedule", "review", "prepare", "check", "define", "align"
• Portuguese equivalents: "acompanhar", "enviar", "agendar", "revisar", "preparar", "verificar", "definir", "alinhar", "vou", "preciso"
• Lines with a verb + future intent: "I'll", "we'll", "we need to", "we should", "needs to"
• Explicit "action:" or "next step:" labels

For each extracted action item:

Determine time horizon using these signals in the text:

Signal	Tag
"today", "hoje", "EOD", "end of day", "amanhã", "tomorrow"	#today
"this week", "essa semana", "by friday", "next sync", "sprint", "semana"	#this-week
"this month", "end of month", "próximo mês", "monthly"	#this-month
"Q1", "Q2", "Q3", "Q4", "quarterly", "next quarter", "trimestre"	#this-quarter
No clear signal	Ask: No timing signal found for: "[task text]". Assign to: (1) today (2) this-week (3) this-month (4) this-quarter [press Enter for this-week]:

Format each task as:

- [ ] [action item text, cleaned up] #[horizon] [[new-filename-without-extension]]

The [[new-filename-without-extension]] is the new filename computed in step 2c, with .md removed. This backlinks the task to the source meeting in Obsidian's graph.

Example:

- [ ] Send AI workshop invite to all team leads #this-week [[2026-04-10-ai-enablement-communication-plan]]
- [ ] Review Q2 objectives before next meeting #today [[[e-mail]]]

Append to my-tasks/tasks.md:

If my-tasks/tasks.md does not exist, create it first with this content:

---
objective: Canonical task list — source of truth for all action items
---

# Tasks

Then append the new tasks section.

Read the current content of my-tasks/tasks.md. Append at the end:

### YYYY-MM-DD — [meeting title from frontmatter]

- [ ] [task 1] #[horizon] [[new-filename-without-extension]]
- [ ] [task 2] #[horizon] [[new-filename-without-extension]]

If no action items were found in the meeting, skip this file — do not append an empty section.

Step 2g: Write updated file to destination and remove from inbox

Write the complete updated meeting note (frontmatter + injected backlinks + original content) to the destination path using the new filename computed in step 2c.

Then delete the original file from inbox/:

rm "inbox/[original-filename]"

Step 2h: Update person profile with meeting backlink

This step applies to all types (1:1, project, company).

For each attendee in the People: line (except Marlon):

1. Find their profile file: [their-folder]/[email].md To resolve [their-folder], search in this order:
   1. my-teams/members/[email]/ — if folder exists, use it
   2. my-leadership/[email]/ — if folder exists, use it
   3. my-company/contractors/[email]/ — if folder exists, use it
   4. my-company/members/[email]/ — if folder exists, use it
2. Check if a ## Meetings section exists at the end
3. If it does NOT exist, append it:

   
   ## Meetings
   
   - [[new-filename-without-extension]]
   

4. If it DOES exist, append a new line inside it:

   - [[new-filename-without-extension]]
   

For project type meetings, also update the project index file if it exists:

• Find my-company/projects/[project-name]/[project-name]-project.md
• If it exists, append to a ## Meetings section the same way

---

Summary

After processing all files, print this summary:

tmr-inbox complete.

Files processed: N
  [original-name] → [new-name]
       destination: [path]
  ...

Profiles created: N
  [email] in [folder]
  ...

Projects scaffolded: N
  [project-name] → my-company/projects/[project-name]/
  ...

Emails inferred: N
  [Full Name] → [inferred-email] (verify)
  ...

Tasks extracted: N total
  #today:        X
  #this-week:    Y
  #this-month:   Z
  #this-quarter: W

All tasks appended to my-tasks/tasks.md
Skipped: [list files the user deferred (option d) and any files skipped for other reasons, or "none"]