Notice each message on this Mac, and tell the engine
Watch the Messages database on the person's own Mac and tell their engine one fact for each iMessage or text message that arrives, is edited or is unsent, the moment Messages writes it, so the workflows that hear messages start. It only reads: it never sends a message.
The noticer's program (tell-the-engine.mjs) run on the person's own Mac with one capability, read-messages, which the person grants to this tool and which macOS allows only once the app that runs the engine has Full Disk Access. Its one wake, keep-noticing-one-source, names no waiter: the person switches it on with their own input (wake:switch input {waiters}), naming the imessage-chat-database leaf, so which Messages database is watched and from when (history_since) is their own DATA. Each message is told as a something-changed fact at place message, with the same eight fields every message source tells (where, place, change, at, key, by, after, kinds), so the router and the history keeper take it as they take WhatsApp. A Mac that refuses the database is told once at place access with the setting to change.
What it asks for
Nothing: your AI starts it with your words.
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to watch the Messages database on the person's own Mac and tell their engine one fact for each iMessage or text message that arrives, is edited or is unsent, the moment Messages writes it, so the workflows that hear messages start. It only reads: it never sends a message”