---
title: "Manuscripts that stay useful"
description: "Use when you keep a manuscript that agents add to over time. Covers where learnings go first, the log when line, approval and auto-update per manuscript, and moving entries into chapters."
url: https://docs.heytrove.ai/best-practices/manuscripts-that-stay-useful
updated: 2026-10-06
---

# Manuscripts that stay useful

> Load this page when: you need to keep a manuscript useful over time

## For agents

- At the end of a task, compare what you learned with each `log when:` line.
- On a match, run `trove ms show <id>` and follow its instructions.
- Auto-update on means write, then report in one line. Auto-update off means propose and wait.
- Stay silent when no manuscript matches.

## Summary

Put what you learn into a [manuscript](https://docs.heytrove.ai/concepts/manuscripts.md) first. Each manuscript has its own approval setting, and over time you move the best entries into chapters.

Want your agent to follow these practices? Ask it: "Read How to Use Trove in my Trove library and follow it."

## Keep your learnings in a manuscript

A manuscript is the place your next agent looks. A learning left in a chat is gone when the session ends. A learning in a repo file or a ticket is hard to find later.

Good learnings are decisions, pitfalls, how something really behaves, and rules you state when you correct your agent's work. When a task ends, ask your agent to log them:

```text
Log what we learned in this task to my Trove manuscripts.
```

Do this explicitly. Trove has automatic checks, and we are actively working on making them more reliable, but today you should not rely on your agent to remember. When the learning matters, ask.

## Choose the approval setting for each manuscript

You set the rule for each manuscript. There is no global rule. Many people keep approval on while a book is still taking shape.

- With auto-update on, the agent writes the entry and tells you in one line.
- With auto-update off, the agent shows you the entry and waits for your yes.

To turn auto-update on for one manuscript, ask your agent:

```text
Turn on auto-update for my "Billing decisions" manuscript in Trove.
```

## Write a sharp "log when" line

Each manuscript has a `log when:` line. It tells the agent what belongs in this manuscript. A vague line, such as "anything useful", matches too much or too little.

Name the kind of learning and the topic, for example "a billing decision and its reason". Add instructions that say how to write an entry, for example "one dated entry per decision, current state first".

```text
Create a Trove manuscript named "Billing decisions". Log when: a billing decision and its reason. Instructions: one dated entry per decision, current state first.
```

## Do not force a learning into the wrong manuscript

If no manuscript fits, a forced entry only makes the wrong manuscript messy. Make a new manuscript only when the topic will clearly come back. Do not make one for a one-time learning. You can ask your agent to create a new manuscript with a clear `log when:` line.

## Move the best entries into chapters

A manuscript grows by entries. A reader needs chapters. From time to time, move the best entries into proper chapters. You can do this yourself, or ask your agent to do it and approve the result.

When you state a general rule while you correct your agent, you can ask it to add the rule to the matching book. Check the wording before it goes in, unless that manuscript allows direct updates.

## Direct CLI

Your agent runs these through the `trove` command. You can run them yourself in a terminal, but asking your agent is the recommended way: it picks the right flags, waits for processing and checks the result.

| Task | Command |
|---|---|
| Turn on auto-update | `trove ms update <id> --auto-update` |
| Create a manuscript | `trove ms create "Billing decisions" --criteria "a billing decision and its reason" --instructions "One dated entry per decision. Current state first."` |
| See a manuscript and its rules | `trove ms show <id>` |

See the [CLI reference](https://docs.heytrove.ai/reference/cli.md).

## Related

- [Manuscripts](https://docs.heytrove.ai/concepts/manuscripts.md) explains the idea.
- [Log what you learn](https://docs.heytrove.ai/usage/log-what-you-learn.md) shows the steps.
- [trove ms](https://docs.heytrove.ai/reference/cli/ms.md) lists the commands.
- [Write books agents can read](https://docs.heytrove.ai/best-practices/write-books-agents-can-read.md) covers chapter structure.