chronicle-email

name	chronicle-email
description	Indexes MS365 Outlook email in two modes: monthly-index writes a full-month markdown index, and recent-window summarizes the last N days for chronicle-day-plan. Use for /chronicle-email, chronicle emails, last week's emails, or summarize this week's inbox.

Chronicle Email

Index emails from MS365 Outlook ([e-mail]) via the Softeria ms-365-mcp-server MCP. Two modes — pick by user intent.

mode = monthly-index   →  full month → markdown index file        (existing)
mode = recent-window   →  last N days → compact summary in-chat   (NEW)
                          (used by chronicle-day-plan Phase 1)

Server: ms-365-mcp-server (Softeria), launched with --preset mail --read-only. Tool namespace: mcp__ms365__*. Token cached in macOS Keychain after one-time device-code login (mcp__ms365__login → phone OAuth).

Output locations:

• monthly-index: ~/Desktop/jluo41-repo/3-Chronile/Emails/MonthIndex/
• recent-window: returned in-chat (no file written by default)

Mode: recent-window (last N days)

Pulls the last N days (default 7) of inbox messages and summarizes in-chat. The PRIMARY consumer is chronicle-day-plan Phase 1 — it asks "anything overnight needing a real reply?" and uses this as fuel.

Verified call (confirmed working 2026-04-28)

tool:    mcp__ms365__list-mail-messages
filter:  receivedDateTime ge {YYYY-MM-DD}T00:00:00Z   ← last N days = today − N
orderby: receivedDateTime desc
select:  id,subject,from,receivedDateTime,bodyPreview,isRead,hasAttachments
top:     15            ← start small; raise to 30 only if days are dense

⚠️ Microsoft Graph rule: do not combine $filter and $search. Use one or the other. For date-window pulls, $filter is correct.

Response shape (what to expect)

{
  "value": [
    {
      "id": "...",
      "receivedDateTime": "2026-04-28T08:34:19Z",
      "subject": "Re: ...",
      "bodyPreview": "...",
      "isRead": true,
      "hasAttachments": false,
      "from": { "emailAddress": { "name": "...", "address": "..." } }
    }
  ],
  "@odata.nextLink": "..."   // present iff more pages
}

Pagination

For >15 emails, follow @odata.nextLink ONCE; refuse to fetch all pages unless the user explicitly says so (avoid huge context blobs). Or set fetchAllPages: true to merge — but only when user wants a complete export, not a daily summary.

Summarization rules (the part that goes back to the caller)

For each message, classify into ONE of:
  ⭐ needs-reply    real correspondence; from a known human; awaits response
  📋 fyi-internal   meeting cancellations, schedule notices, internal admin
  💸 financial      invoices, receipts, billing — flag with [$]
  📰 newsletter     bulk mail, marketing, digest — drop unless explicit ask
  🔒 security       account-security notices — drop unless they're flagging risk

Return groups in priority order (needs-reply → fyi-internal → financial),
with newsletters/security counted but not enumerated.

Compact in-chat output template

📨 Last 7 days · N emails · M after filter

⭐ NEEDS REPLY (count)
  · 04-28  Ed Bunker         — Re: About the Dissertation Proposal
  · 04-28  Heller, Carol     — Re: Wharton poster prep
  · 04-27  Susanne Muehlsch. — Undergrad student project inclusion?
  · 04-27  attorneys@chen…   — EB-1A Evaluation

📋 FYI / INTERNAL (count)
  · 04-27  Minghong Xu       — AI Agent Lab weekly meeting (cancelled)

💸 FINANCIAL (count)
  · (none worth flagging)

(N newsletters · M security notices — dropped)

This compact form is what chronicle-day-plan Phase 1 uses as a SEED for sharp prompts ("you've got 4 emails awaiting real replies — which one needs your brain today?"). It is not the plan output itself.

Anti-patterns

• ❌ Returning raw bodyPreview unfiltered → too noisy for a day plan
• ❌ Including "External Email" warning lines in summaries → strip them
• ❌ Pulling more than 15 emails for the day-plan use case → context bloat
• ❌ Persisting to a markdown file → recent-window is in-chat only; monthly-index is the persistent mode

---

Mode: monthly-index (full month → markdown file)

Index Format

Monthly email indexes follow a standardized format with:

• Header: Total count, period coverage, reimbursement note
• Pre-month section: Older email threads filed into the month folder
• Daily sections: Organized by date with table format (Date | From | Subject)
• Reimbursement summary: Expense-related emails at the end

Entry Format Example

# 2026-01 Email Index

**Total**: 604 emails in folder `2026_01`
**Period**: Full month (Jan 1-31, 2026)

> Emails marked with **[$]** are reimbursement/receipt/expense-related.

---

## Pre-January (older threads filed into this folder)

| Date | From | Subject |
|------|------|---------|
| 2025-09-23 | Gordon Gao | Re: HBR Pitch on the Collapse and Future of Telehealth Platforms (x7) |
| 2025-12-17 | Risa Wolf | Re: Clinical-decision support for AID systems (x2) |

---

## Jan 1

| Date | From | Subject |
|------|------|---------|
| 01-01 | LinkedIn | You appeared in 30 searches |
| 01-01 | **[$]** Amazon Web Services | **Amazon Web Services Invoice Available [Account: 415909850145]** |

## Jan 2

| Date | From | Subject |
|------|------|---------|
| 01-02 | BlueStar | Let's get back on track: your weekly challenge |
| 01-02 | New York Times | Inside the unraveling U.S.-Ukraine partnership |

---

## Reimbursement & Expenses Summary

| Date | From | Subject | Notes |
|------|------|---------|-------|
| 2026-01-01 | Amazon Web Services | Amazon Web Services Invoice Available [Account: 415909850145] | AWS monthly invoice |
| 2026-01-15 | [e-mail] | [sis-bill-notification] JHU Monthly eBill Broadcast | JHU monthly billing statement |

Implementation Steps

1. Parameter Gathering

• Required: Month (e.g., "2026-01", "January 2026", or "2026_01")
• Default: Use previous month if not specified

2. MS365 Connection

Use mcp__ms365__login to authenticate
Use mcp__ms365__verify-login to check status

3. Locate Email Folder

Navigate folder structure to find the target month:

Inbox → jluo41-{YEAR} → {YEAR}_{MONTH}
Example: Inbox → jluo41-2026 → 2026_01

Folder naming convention:

• Parent: jluo41-{YEAR} (e.g., "jluo41-2026")
• Monthly: {YEAR}_{MONTH} (e.g., "2026_01")

4. Retrieve All Emails

Use mcp__ms365__list-mail-folder-messages with:

• mailFolderId: The folder ID from step 3
• select: "subject,from,receivedDateTime" (comma-separated string)
• orderby: "receivedDateTime asc" (single string)
• fetchAllPages: true

Important: Handle pagination properly. Monthly folders typically contain 300-600+ emails.

Server constraint: Cannot combine $filter and $search in the same call. For monthly-index, no filter is needed (the folder is already scoped); for date-windowed pulls (recent-window mode), use $filter only.

5. Process and Format Emails

Grouping:

• Separate pre-month emails (before first day of target month)
• Group by day for the target month
• Track duplicates (same sender + same subject on same day)

Date Formatting:

• Pre-month: Use YYYY-MM-DD format
• Target month: Use MM-DD format for daily sections
• Section headers: Use "Jan 1", "Feb 15" style (month abbreviation + day)

Reimbursement Detection: Mark emails with [$] if subject or sender contains keywords:

• Financial: invoice, receipt, billing, payment, order confirmation, reimbursement
• Vendors: Amazon Web Services, AWS, Uber, cybersource, expense
• JHU: ebill, billing statement

Duplicate Handling: When same sender sends multiple emails with identical subject on same day:

• First occurrence: Show normally
• Subsequent: Add (x2), (x3), etc. suffix to subject

6. Write Output File

File path: ~/Desktop/jluo41-repo/3-Chronile/Emails/MonthIndex/{YEAR}-{MONTH}-index.md

File structure:

1. Header with total count and period
2. Reimbursement note
3. Horizontal rule ---
4. Pre-month section (if applicable)
5. Horizontal rule ---
6. Daily sections for each day with emails
7. Horizontal rule ---
8. Reimbursement & Expenses Summary

7. Context Optimization

To save context tokens:

• Use Task/general-purpose agent for email retrieval and processing
• Process data in background agents when dealing with 500+ emails
• Only load final formatted output into main conversation

Writing Style Guidelines

Subject Lines:

• Preserve original subject exactly as received
• Bold entire subject line for reimbursement emails
• Add duplicate count suffix when applicable

Sender Names:

• Use from.emailAddress.name field (display name)
• If display name is empty, use email address prefix

Table Formatting:

• Keep columns aligned with pipes |
• Use consistent spacing
• No line breaks within table rows

Error Handling

• If MS365 not logged in, authenticate first
• If folder not found, check folder structure and year
• If no emails in period, create minimal index noting "0 emails"
• If partial month, note date range in period field

Reimbursement Summary Guidelines

For each expense-related email, provide:

• Date: Full YYYY-MM-DD format
• From: Sender name or organization
• Subject: Full subject line (without [$] markers)
• Notes: Brief categorization (e.g., "AWS monthly invoice", "Uber ride receipt")

Usage Examples

User request: "Chronicle emails for January 2026" Action:

1. Connect to MS365
2. Find 2026_01 folder
3. Retrieve all emails
4. Generate index at 3-Chronile/Emails/MonthIndex/2026-01-index.md

User request: "/chronicle-email 2025-12" Action: Process December 2025 emails from 2025_12 folder

User request: "Update my email chronicle for last month" Action: Auto-detect previous month and process accordingly

Tags Strategy

No tags needed for this skill (unlike logseq-cc-records). The Chronicle system uses folder-based organization.

Output Confirmation

After successful indexing, report to user:

• Total emails indexed
• Number of expense-related emails
• File location
• Date range covered