recent-conversations

name	recent-conversations
description	Summarise your most recent message exchanges — who you've been talking to, what topics came up, and the overall tone of recent activity.

Recent Conversations

Summarise what's been happening in your messages lately.

When to Use

• "What have I been talking about recently?"
• "Who have I been messaging this week?"
• "Catch me up on my recent conversations"
• "What's been going on in my messages lately?"

Steps

1 — Check and sync

status()

If last_sync is more than a few hours old, call sync() first.

2 — Get recent activity window

Ask the user how far back to look if not specified. Default to 7 days.

-- Most active contacts in the last 7 days
SELECT
  name,
  COUNT(*) AS messages,
  MAX(CAST(timestamp AS VARCHAR)) AS last_message,
  SUM(CASE WHEN is_from_me = 1 THEN 1 ELSE 0 END) AS sent,
  SUM(CASE WHEN is_from_me = 0 THEN 1 ELSE 0 END) AS received
FROM messages
WHERE timestamp >= CAST(date_trunc('day', now() - INTERVAL '7 days') AS TIMESTAMP)
  AND name IS NOT NULL
  AND chat_size = 1
GROUP BY name
ORDER BY last_message DESC
LIMIT 15

-- Recent group chat activity
SELECT
  chat_members_contact_info,
  COUNT(*) AS messages,
  MAX(CAST(timestamp AS VARCHAR)) AS last_message
FROM messages
WHERE timestamp >= CAST(date_trunc('day', now() - INTERVAL '7 days') AS TIMESTAMP)
  AND chat_size > 1
GROUP BY chat_members_contact_info
ORDER BY last_message DESC
LIMIT 5

-- Recent message volume by day
SELECT
  CAST(date AS VARCHAR) AS day,
  COUNT(*) AS total,
  SUM(CASE WHEN is_from_me = 1 THEN 1 ELSE 0 END) AS sent,
  SUM(CASE WHEN is_from_me = 0 THEN 1 ELSE 0 END) AS received
FROM messages
WHERE date >= date_trunc('day', now() - INTERVAL '7 days')
GROUP BY date
ORDER BY date DESC

3 — Summarise

Give a brief narrative: who's been most active, any notable conversations (high volume days, new contacts), and your overall send/receive balance for the period. Keep it concise — this is a catch-up, not a deep analysis.

Offer to do a contact-deep-dive on anyone who looks interesting in the results.

Note on message content

This skill works entirely from metadata (names, counts, timestamps). It does not read the text of your messages.