| name | email_briefing |
| description | Natural-language summary of recent email activity across all of the user's accounts |
| when_to_use | - User asks "what's new in my email?" / "any important emails today?"
- Morning routine: combined with whatsapp_briefing for a full inbox overview
- Before composing a reply, agent wants a quick read of the email landscape
- For finding a SPECIFIC email by sender or subject phrase, call `find_email_by_subject` first — this skill is for broad recent-activity summaries only.
|
| inputs | {"hours":{"type":"integer","required":false,"default":24,"description":"Look-back window in hours (1-168)"}} |
| outputs | {"summary":{"type":"string","description":"Markdown briefing — \"Action items\" + \"By thread\" sections, auto-language matched to the inbox content"},"stats":{"type":"object","description":"Object with keys hours / accounts / new_messages / threads_needing_reply / unread_count"},"threads_needing_reply":{"type":"array","description":"List of {message_id, thread_id, from, subject} for messages the user hasn't replied to"},"generated_at":{"type":"string"}} |
| cost | 1 LLM call (~3-8s on qwen3) |
| permissions | ["admin","member"] |
| side_effects | none — read-only |
| tags | ["email","summary","llm"] |
email_briefing
Aggregates recent inbound email from all of the user's accounts grouped
by sender/thread, identifies which ones plausibly need a reply (you
got an email and haven't sent one in the same thread yet), and asks
the LLM for a prioritised briefing.
Auto-language: prompt tells the LLM to write in German if most messages
are German, English otherwise.
Naturally pairs with whatsapp_briefing — a future "morning_briefing"
composite skill calls both and stitches them.
summary is user-ready markdown — surface it verbatim or with one framing sentence, and never re-summarise it or enumerate threads_needing_reply in prose.
