You ask for a fix and Verboo Code hands back finished code without explaining the choice. Sometimes that is exactly what you want. Other times you wanted to understand why, or you wanted the agent to stop and let you write the hard part yourself. That is what an output style changes: the way the coding agent talks to you, without touching what it knows how to do.
Where did /output-style go?
It was deprecated. Type /output-style in Verboo Code and you get this message:
/output-style has been deprecated. Use /config to change your output style, or set it in your settings file. Changes take effect on the next session.
Output styles still exist. Only the place to pick them changed: now it is /config or the settings file, and the change applies from the next session.
Which output styles come built in?
Three. The default one and two aimed at learning while the agent works:
| Style | What changes in the answer | Good for |
|---|---|---|
Default | Completes the task efficiently and answers concisely | Everyday work |
Explanatory | Before and after writing code, adds a ★ Insight block with 2 or 3 points about implementation choices and the project's patterns | Getting into code you don't know |
Learning | Everything in Explanatory and, when the task goes past 20 lines with a design decision, leaves a TODO(human) in the code and asks you to write 2 to 10 lines | Hands-on practice |
In Learning, the agent stops after the Learn by Doing request and waits for you to implement before moving on. If you just want the task done, skip this one.
How do I change the output style with /config?
Inside the session:
/config
Select Output style. The Preferred output style panel opens with the three built-in styles and the ones you created. Pick one and leave /config.
The choice is saved to .verboo/settings.local.json, in the project folder, under the outputStyle key. Close the session and open a new one:
verboo
From then on, the agent answers in the new style.
Can I set the style straight in the settings file?
Yes. The key is outputStyle and the value is the style name:
{
"outputStyle": "Explanatory"
}
The file you put it in decides who gets the style:
| File | Applies to |
|---|---|
~/.verboo/settings.json | You, in every project |
.verboo/settings.json | Everyone who clones the repository (goes into git) |
.verboo/settings.local.json | You, in this project only (where /config writes) |
To go back to the default, use "outputStyle": "default" or delete the key.
Trap: the name has to match exactly, including upper and lower case. "outputStyle": "explanatory" does not find the Explanatory style, and Verboo Code does not warn you: it silently falls back to the default. If the style didn't take, check the spelling first.
How do I create my own output style?
One Markdown file per style, in an output-styles folder. The file content becomes the style's instructions. Here is a style that asks the agent to explain the risk before editing:
mkdir -p .verboo/output-styles
And the file .verboo/output-styles/revisor.md:
---
name: Revisor
description: Points out the risk of each change before editing and keeps answers short
keep-coding-instructions: true
---
# Revisor style
- Before editing any file, list in up to 3 bullets what will change and the risk of each change.
- Answer in short sentences, without repeating the code you just wrote.
- When you finish, say which test command to run to check the change.
Open /config and Revisor is already on the list, with the description you wrote. The three frontmatter fields:
| Field | If you leave it out |
|---|---|
name | Uses the file name without .md (revisor) |
description | Takes the description from the file content itself |
keep-coding-instructions | Stays off (see the next section) |
outputStyle takes the name, not the file name. With the example above, the value is "Revisor".
Why did the agent stop coding properly with my style?
Because keep-coding-instructions: true is missing from the frontmatter. Without that field, a custom style removes from the system prompt the section on how to carry out coding tasks, and the agent is guided by your style alone.
The built-in Explanatory and Learning styles already have the field on, which is why they keep the coding behavior. Your own styles don't: if the style is meant for coding work, write the field. Leaving it out only makes sense when you want the agent for something else, like writing docs or answering questions.
Where does Verboo Code look for styles?
| Folder | Scope |
|---|---|
~/.verboo/output-styles/ | Yours, in every project |
.verboo/output-styles/ or .claude/output-styles/ | The project's. Read from the current folder up to the git repository root |
The .claude/output-styles/ folder is read too, so project styles you already kept there work without copying. The user folder, though, is only ~/.verboo/output-styles/: a style in ~/.claude/output-styles/ does not show up. If you moved the config folder with VERBOO_CONFIG_DIR, the user folder becomes $VERBOO_CONFIG_DIR/output-styles/.
When two styles share a name, the more specific scope wins: the project's beats yours, and yours beats the built-in one. An Explanatory.md with name: Explanatory in the project replaces the built-in style for everyone working in it.
Output style or AGENTS.md: which one?
Each one changes something different:
| Output style | AGENTS.md / /memory | |
|---|---|---|
| Changes | How the agent talks: length, format, whether it explains, whether it asks for your help | What the agent knows about the project: commands, conventions, where things live |
| One at a time? | Yes, only one active style | No, the files add up |
"Run pnpm test before saying you're done" is a project rule and belongs in AGENTS.md, as in the tutorial on permanent instructions with /memory. "Answer in bullets and explain the risk" is a way of talking and belongs in the output style.
Explanatory and Learning make answers longer and sessions more back-and-forth. On Verboo Code tokens are unlimited, so the extra explanation doesn't turn into an extra bill.
Behavior verified in the Verboo Code source (verbeux-ai/code, v0.15.41) on 2026-11-04: src/commands/output-style, src/constants/outputStyles.ts, src/outputStyles/loadOutputStylesDir.ts, src/constants/prompts.ts, src/components/Settings/Config.tsx and src/utils/markdownConfigLoader.ts.



