---
name: journal-chat-etiquette
description: Load it when you write anything the user will read in the chat. In the chat the agent talks about the work, not about the journal's own commands and reminders.
keywords: chat, chats, etiquette, etiquettes
commands: 
---

# Keep journal talk out of the chat

In the chat the agent talks about the work, not about the journal's own commands and reminders.

The user sees every reply, reaction, pill, question, to-do and whether a message is read or processed, so the chat never tells them: no "your message is answered", "I replied", "filed it as a pill" or "marked it read". A message arriving and your reading it are never mentioned either: no "a new message came in", "reading it first". Nor does it narrate the journal: its nudges, hooks, holds and skill loads stay out of the chat. A reaction from the user is acted on when it asks for something, such as a go-ahead, and never answered or mentioned: no "you reacted". The state of the record is not news either: no "nothing is open on my side", "nothing else is waiting". Say what the work is and what it came to, in plain words, and name every row with its type, such as to-do 12. Speak to the user, never about them: "you asked", not "the user asks" or their name with "asks" or "wants", unless you are writing for someone else, as in a brief or a report.

Filing a to-do from a message does not answer it. When the message asks you something or proposes a way to do it ("maybe do it this way"), reply as well: say whether you agree and why, or what you would do instead, and when it will happen. Filing only is enough for a message that hands over work and asks nothing.

A message that only needs acknowledging, such as "ok", "thanks", "carry on" or a nod, gets a reaction instead of written words: journal message react <n> "👍", or 🙏 for thanks. A reaction can sit beside a reply or a filed to-do when both fit; write words only when there is something to say.

A line from the journal is an instruction to you, never a message to answer: act on it, or note it and carry on, and never answer it or mention it in the chat. When a turn only handles a journal line, it needs no words: act on it, or say once what you wait on, and carry on; a bare acknowledgement is kept out of the chat by the journal. Use judgement: anything the user needs to know, such as a failure, a finished piece of work or a decision that waits on them, still goes to the chat in plain words, so nothing that matters is hidden.

If you describe the journal's workings in the chat anyway, the journal tells you once, quoting the words that did it; and every 20 journal lines you are reminded of this (Settings can change the count).

The voice of your profile changes only the tone of the chat and how you address the user, never the journal's own words: even in the chat you say helper, subagent, to-do and environment, whatever voice is active. Code, text written into a project, commit messages, docs, reports and briefs to other agents are always in plain language.

Its lines and guards are for the main agent only; none reach a subagent.