Best Practices

Start Important Work In Plan Mode

For anything non-trivial, start with:

/plan <task>

This gives you a structured proposal before YeePilot starts taking action.

Keep Safety Tight On Important Machines

Recommended baseline for production-sensitive work:

security:
  mode: strict
  require_confirmation: true
 
ai:
  autonomy_profile: strict_review

Keep Sessions Focused

Use separate sessions for separate goals:

• one session for one incident
• one session for one repo task
• one session for one server audit

That makes results easier to review later and keeps context cleaner for the model.

Attach Real Files Early

Use /file or /upload as soon as the task depends on a real file. It is better than asking YeePilot to guess your config or project structure.

Watch Live Changes

Keep the live stream panel enabled for mutating tasks:

tui:
  live_stream_panel_enabled: true

It is one of the easiest ways to stay oriented while YeePilot edits files.

Use The Neural HUD Instead Of Memorizing Everything

The HUD is the fastest place to:

• switch sessions
• inspect usage and plan access
• review audit activity
• change settings
• customize keybinds

Use The Vault Before You Need It Under Pressure

Set up the vault in advance, not during an outage:

yeepilot vault init --method password

Also:

• store the recovery key offline
• verify SSH host fingerprints
• keep autolock enabled

Manage Cost Without Losing Control

If sessions are getting long:

• switch to /tokenmode saver
• lower thinking when the task is simple
• remove old file attachments
• start a new session instead of piling several unrelated tasks into one

Keep The Client Updated

Check periodically:

yeepilot update --check
yeepilot update

If you are invited to test a protected canary build, use the canary channel intentionally and keep the token private.

Report Problems Cleanly

When something genuinely breaks, do not rely on memory alone. Use:

yeepilot support report --message "Describe the problem"
yeepilot diagnostics pending

This gives support and operations a much better starting point than a vague summary.