Guide Editor
The Guide Editor provides a comprehensive interface for building and modifying guides. This guide covers the editor's features and how to construct effective procedural automation.
Opening the Editor
Access the Guide Editor by:
- Creating a new guide (click New on the Guides page)
- Editing an existing guide (click Edit on a guide's detail page)
- Opening a draft generated from Chat
Editor Layout
The editor provides several areas:
| Section | Purpose |
|---|---|
| Header | Title, description, metadata |
| Steps Panel | Add, edit, and order steps |
| Variables Tab | Configure input and output variables |
| Test Run | Validate the guide with sample data |
Guide Metadata
At the top, configure basic information:
- Title — Descriptive name for the guide
- Description — What this guide does and when to use it
- Category — Grouping for organization
- Tags — Keywords for search
Working with Steps
Steps are the building blocks of guides. Each step represents one action in the procedure.
Adding Steps
- Click Add Step
- Enter a title and description
- Configure the step type and automation
- Save
Step Configuration
For each step, configure:
| Field | Description |
|---|---|
| Title | Brief name (e.g., "Check database status") |
| Description | Detailed instructions for what this step does |
| Automation Type | How the step executes |
| Tool Configuration | For automated steps, which tools to use |
| Success Indicator | How to verify the step completed correctly |
Automation Types
| Type | Behavior |
|---|---|
| Fully Automated | AI executes using tools, no human involvement |
| Semi-Automated | AI executes with human confirmation before each action |
| Manual | Human performs the action and marks complete |
| Information Only | Displays information, no action required |
Configuring Automated Steps
For automated and semi-automated steps:
- Select Tools — Choose which tools from your outposts can be used
- Provide Instructions — Tell the AI what to accomplish
- Set Context — Specify what information the step needs
- Define Output — What data to capture from the step
The AI interprets your instructions, selects appropriate tools, executes them, and interprets the results.
Manual Steps
For manual steps:
- Write Clear Instructions — What the human should do
- Add Command Blocks — Copy-able commands if applicable
- Specify Expected Outcome — What success looks like
- Include Verification — How to confirm the step worked
Command Blocks
For steps that involve commands:
command_block:
title: "Check service status"
command: "systemctl status myservice"
copy_enabled: true
Command blocks render with a copy button, reducing errors from manual typing.
Step Order
Steps execute in the order shown. To reorder:
- Drag and drop steps in the editor
- Use up/down arrows on each step
Variables
Variables make guides flexible and reusable.
Input Variables
Input variables are provided when the guide starts.
Adding an Input Variable:
- Go to the Variables tab
- Click Add Input Variable
- Configure:
- Name — Reference name (e.g.,
server_name) - Type — Text, number, boolean, selection
- Required — Must be provided to run
- Default — Value if not specified
- Description — Help text for users
- Name — Reference name (e.g.,
Example Input Variables:
environment— dev, staging, prodserver_name— Target serverbackup_retention_days— How long to keep backups
Output Variables
Output variables capture results for later steps or final output.
Adding an Output Variable:
- Click Add Output Variable
- Configure:
- Name — Reference name
- Type — Expected data type
- Source Step — Which step produces this value
Using Variables in Steps
Reference variables in step instructions:
Connect to server {{server_name}} in the {{environment}} environment.
The editor provides autocomplete for available variables.
Passing Data Between Steps
Automated steps can capture output that later steps use:
- Step 1 runs a query, captures result as
db_status - Step 2 references
{{db_status}}in its instructions - AI uses the previous result to inform the current step
Test Run
Validate your guide before publishing.
Running a Test
- Go to the Test Run tab
- Provide values for input variables
- Click Start Test
- Watch execution progress
- Review results for each step
Test Mode Behavior
In test mode:
- Tools execute normally (be careful with destructive actions)
- Results are logged but don't count in metrics
- You can step through manually or let it run
Debugging
If a test fails:
- Review which step failed
- Check tool execution logs
- Verify variable values
- Adjust configuration and retest
Publishing
Before Publishing
Verify:
- All steps have clear descriptions
- Automated steps have correct tool configuration
- Variables are properly defined
- Test run completed successfully
Publishing the Guide
- Click Publish
- Confirm you want to make it available
- Guide state changes from Draft to Active
Published guides appear in the guide library for your team.
Best Practices
Let AI generate a first draft from your description, then refine in the editor.
Each step should do one thing. Break complex actions into multiple steps for clarity and easier debugging.
"Rotate database credentials" is better than "Step 3". Titles should explain what happens.
Use variables instead of hardcoded values. A guide that works for one server should work for any server.
For critical steps, add verification or rollback instructions in case of failure.
Note prerequisites, required permissions, and expected environment in the description.
Related Topics
- Guides — Guide concepts and overview
- Flows — Diagnostic automation
- Flow Editor — Similar concepts for flows
- Runs — Track guide executions