[FEATURE] Option to set output style in non-interactive mode

[j-greig]

Description

Currently (v1.0.81) output styles as introduced in the recent release only work in interactive mode via /output-style command. There's no way to use custom output styles with the -p (print/headless) mode, which limits automation and scripting capabilities.

Current Behavior

• Output styles can only be set interactively using /output-style [style]
• The outputStyle field in .claude/settings.local.json doesn't affect -p
  mode
• Custom output styles in .claude/output-styles/ cannot be used with headless
  execution

Desired Behavior

Add support for output styles in non-interactive mode through one or more of
these approaches:

1. CLI flag: claude -p "prompt" --output-style french-teacher
2. Settings respect: Have -p mode respect the outputStyle field in
   settings.json
3. System prompt integration: Allow output style files to be loaded as system
   prompts

Use Cases

• Automated generation with specific communication styles (e.g., different
  languages, educational modes)
• Batch processing with custom output requirements
• Integration with tools expecting specific output formats

[github-actions]

Found 1 possible duplicate issue:

1. [BUG] Documentation inconsistency: outputStyle setting documented but not supported #6126

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

[coygeek]

Hey, great feature request. I've been wanting to do the same thing for some scripting I'm working on. A dedicated --output-style flag would definitely be the ideal solution.

In the meantime, I've found a pretty good workaround using the existing CLI flags mentioned in the documentation. It's not as clean as your proposal, but it gets the job done for automation.

The trick is to use the --append-system-prompt flag, which is specifically designed to work with the -p (print) mode.

Essentially, you can take the core instructions from an output style and pass them directly as a system prompt.

For example, if you wanted to use your french-teacher style, you could do something like this:

claude -p "Explain what a promise is in javascript" --append-system-prompt "You are a French teacher. You are patient, encouraging, and clear. Explain concepts in a simple way, using analogies related to everyday life. Provide examples in both French and English."

This forces Claude to adopt that persona for the non-interactive command.

The changelog for version 1.0.55 also mentions a --system-prompt-file option, which is even better for managing these styles. You can save your output style's prompt into a markdown file and reference it directly.

So you could create a file like ./styles/french-teacher.md:

You are a French teacher. You are patient, encouraging, and clear. Explain concepts in a simple way, using analogies related to everyday life. Provide examples in both French and English.

And then run your command like this:

claude -p "Explain what a promise is in javascript" --system-prompt-file ./styles/french-teacher.md

This lets you manage your different "styles" as a collection of prompt files and use them in your scripts. It's basically implementing your third suggestion ("System prompt integration") with the existing tools.

It's not perfect, and I agree that having -p mode respect the settings.json or have a dedicated flag would be much more streamlined. But hopefully this helps you get your automation working for now

[Lenz-AI77]

Would love to use the SDK if the syntax was available.

[brickfrog]

Hey, great feature request. I've been wanting to do the same thing for some scripting I'm working on. A dedicated --output-style flag would definitely be the ideal solution.

In the meantime, I've found a pretty good workaround using the existing CLI flags mentioned in the documentation. It's not as clean as your proposal, but it gets the job done for automation.

The trick is to use the --append-system-prompt flag, which is specifically designed to work with the -p (print) mode.

Essentially, you can take the core instructions from an output style and pass them directly as a system prompt.

For example, if you wanted to use your french-teacher style, you could do something like this:

claude -p "Explain what a promise is in javascript" --append-system-prompt "You are a French teacher. You are patient, encouraging, and clear. Explain concepts in a simple way, using analogies related to everyday life. Provide examples in both French and English."
This forces Claude to adopt that persona for the non-interactive command.

The changelog for version 1.0.55 also mentions a --system-prompt-file option, which is even better for managing these styles. You can save your output style's prompt into a markdown file and reference it directly.

So you could create a file like ./styles/french-teacher.md:

You are a French teacher. You are patient, encouraging, and clear. Explain concepts in a simple way, using analogies related to everyday life. Provide examples in both French and English.
And then run your command like this:

claude -p "Explain what a promise is in javascript" --system-prompt-file ./styles/french-teacher.md
This lets you manage your different "styles" as a collection of prompt files and use them in your scripts. It's basically implementing your third suggestion ("System prompt integration") with the existing tools.

It's not perfect, and I agree that having -p mode respect the settings.json or have a dedicated flag would be much more streamlined. But hopefully this helps you get your automation working for now

Isn't the point of an output style that it supplants the system prompt? This is basically just the equivalent of having "You are a french teacher" in CLAUDE.md

Also +1 for this, there's a lot of neat things I think you could essentially use claude code as essentially a bootstrap for an app with the ability to overwrite the system prompt on demand

EDIT: okay, I'm dumb, you can just dynmically set this with --settings, since you can change output style there, so you can do:

{
  "outputStyle": "french-teacher"
}

or whichever

[github-actions]

This issue has been inactive for 30 days. If the issue is still occurring, please comment to let us know. Otherwise, this issue will be automatically closed in 30 days for housekeeping purposes.

[github-actions]

This issue has been automatically closed due to 60 days of inactivity. If you're still experiencing this issue, please open a new issue with updated information.

[github-actions]

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.