comms-archive

name	comms-archive
description	Archive Outlook emails and Teams messages as a daily summary in the knowledge base. Requires Microsoft 365 MCP connector.
argument-hint	[days back, e.g. '2' for last 2 days] (default: today)

Skill: comms-archive

$ARGUMENTS

What This Does

Fetches Outlook emails and Teams messages via the Microsoft 365 MCP connector, filters out noise, clusters by topic, and writes a daily summary to a comms/ folder in the current repository.

Prerequisites

• Microsoft 365 MCP connector configured (for outlook_email_search, chat_message_search, read_resource)
• Repository with a CONVENTIONS.md (for frontmatter schema and author initials)

Setup (first run)

If comms/ does not exist yet:

Show this privacy notice first and ask for explicit confirmation before proceeding:

Privacy Notice: This skill fetches emails and chat messages from Microsoft 365 and sends their content to an LLM API for summarization. This means personal data (names, email addresses, message content) of you, your colleagues, and external contacts will be processed by the LLM provider.

Before using this skill, ensure:

• Your organization permits sending workplace communications to third-party AI services
• A Data Processing Agreement (DPA) with the LLM provider is in place if required
• You have informed relevant stakeholders (e.g., DPO, works council) if applicable

The generated summaries (containing names and decisions) will be stored in this git repository. Ensure the repository access controls are appropriate.

If the user confirms, proceed:

1. Create comms/ and comms/sync/
2. Create comms/_index.md with frontmatter describing the folder purpose
3. Create comms/sync/processed-ids.json with empty arrays:

   {"last_sync": null, "processed_emails": [], "processed_teams_messages": []}
   

4. Ask the user which noise filters to apply (offer defaults below) and save them to comms/sync/filter-rules.json

Steps

1. Determine date range

• No argument: last 24 hours
• Numeric argument N: last N days
• Use afterDateTime accordingly

2. Load configuration

• Read author initials from CONVENTIONS.md (the default author: field)
• Read noise filters from comms/sync/filter-rules.json (if exists, otherwise use defaults)
• Read processed IDs from comms/sync/processed-ids.json

3. Fetch communications

• Outlook: outlook_email_search with afterDateTime, limit: 50
• Teams: chat_message_search with query: '*', afterDateTime, limit: 50

If a source returns exactly limit (50) results, more likely exist for that window — note the truncation in the summary so a busy day isn't silently under-archived.

4. Deduplicate

Skip messages whose IDs already appear in processed-ids.json. This file grows unbounded; when writing it back, drop IDs older than ~30 days to keep it small.

5. Filter noise

Apply rules from comms/sync/filter-rules.json. Default rules if no config exists:

{
  "exclude_senders": [],
  "exclude_subject_contains": [
    "invited to share this calendar",
    "[MASSENMAIL]"
  ],
  "exclude_subject_starts_with": [
    "Accepted:",
    "Declined:"
  ],
  "exclude_sender_and_subject": [
    {"sender": "noreply", "subject": "Verifying"}
  ],
  "exclude_sender_domains": []
}

Suggest additional filters to the user based on what looks like noise in the fetched messages.

6. Read relevant emails

For each relevant email, use read_resource to get full body. Strip HTML signatures, legal disclaimers, and meeting join links.

7. Cluster and write summary

Group messages by topic/theme. Write to comms/YYYY-MM-DD-daily.md.

Frontmatter (use author initials from CONVENTIONS.md):

---
title: "Daily Communications Summary YYYY-MM-DD"
type: summary
date: YYYY-MM-DD
author: <initials from CONVENTIONS.md>
status: accepted
tags: [daily-summary, <topic-tags>]
---

Section format:

## Topic Name (HIGH PRIORITY if critical)

**Sources:** Teams chat with X, Email from Y

Key facts, decisions, context.

### Action Items
- Who does what

### Upcoming Meetings
- Date/time and purpose

Low-value items go in ## Misc / Low Priority at the bottom.

8. Update tracking

Add newly processed IDs to comms/sync/processed-ids.json. Update last_sync.

9. Confirm and commit

Show the user a brief overview of the topics found. Ask for confirmation before writing. Commit with: docs(comms): daily summary YYYY-MM-DD

Rules

• Respond and write summaries in the user's language (detect from conversation)
• Be concise, capture decisions, action items, deadlines
• Flag HIGH PRIORITY items prominently
• If no new messages: notify user, do not create an empty file
• Never include email signatures, legal disclaimers, or meeting join links
• Always confirm before committing