> ## Documentation Index
> Fetch the complete documentation index at: https://docs.turen.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompts and follow-ups

> Submit work, attach context, steer, queue, stop, and revert

## Compose a prompt

The composer supports text, commands, skills, images, and supported file attachments. Choose the active agent and model before submitting. The first prompt promotes a draft into a durable session.

Attachments remain ordered and session-scoped. Large or unsupported files are rejected before submission.

## Follow-up delivery

TurenOS uses explicit delivery semantics:

| Delivery | Behavior                                                                    |
| -------- | --------------------------------------------------------------------------- |
| Steer    | Promotes at the next safe provider-turn boundary and redirects current work |
| Queue    | Promotes one input at a time after in-flight tool calls settle              |

Steers take precedence over queued input. Promoting new user input resets the selected agent's provider-turn allowance.

## Commands and skills

Slash commands can build structured prompts or invoke configured workflows. Skills load bounded instructions and resources for specialized work. The composer previews optimistic user messages while durable admission is in progress.

## Stop and revert

**Stop** interrupts active local execution without deleting the transcript. **Revert** restores the conversation and workspace to a supported earlier point. Review pending tool activity before reverting.

## Prompting for evidence

Ask the agent to include concrete evidence:

```text theme={null}
Identify the root cause, cite exact files and lines, implement the smallest fix, and report the commands and results used to verify it.
```