Train Korey to Write and Work Your Way

Every team works differently. Some prefer terse, bullet-driven output. Others want narrative descriptions with full context. Some engineers use Cursor; others are on Devin. Korey is designed to adapt to how you actually work — not the other way around.

Here’s how to make it yours.

Adjust writing style in preferences

Head to Settings → Preferences to find two quick wins:

Level of Detail — Choose between Concise, Standard, or Descriptive. Concise keeps responses tight and scannable. Descriptive gives you fuller context, rationale, and explanations. Standard sits in the middle.

Tone — Pick Casual, Neutral, or Formal depending on your team’s communication culture. If your Slack is full of emojis and shorthand, Casual will feel more natural. If you’re writing specs that go into customer-facing documentation, Formal keeps things polished.

Shape output with custom instructions

The Output Format field gives you free-form control over how Korey structures its responses. Some examples of what teams add here:

• Acceptance Criteria required on every story
• Avoid user personas — write from a system perspective
• Always include a Definition of Done section
• Use Given/When/Then format for acceptance criteria

Think of it as a persistent system prompt that applies to everything Korey generates.

Use Memories to store persistent context

Korey’s Memories feature (Settings → Memories) lets you save information that persists across every conversation. This is different from a one-off instruction — once saved, Korey references it automatically.

Useful things to store in memory:

• Your timezone and working hours
• The language your codebase is written in
• The naming conventions your team uses for stories and branches
• Which coding agent you’re handing off to (Cursor, Devin, Jules, etc.)
• Specific project context like stack, architecture decisions, or team structure

Tell Korey directly in a message: “Remember that we write acceptance criteria in Given/When/Then format and always assign stories to the current sprint.” Korey will confirm and store it.

Train custom workflow commands

This is where things get powerful. You can teach Korey a keyword that triggers a multi-step workflow — and from then on, just use that keyword.

Here’s an example: one team trained a command called LIFTOFF. When they type it with a task description, Korey automatically:

1. Creates a Shortcut story from the context provided
2. Assigns it to the current user
3. Initiates a Cursor coding session with the spec attached
4. Generates an initial pull request
5. If it’s a bug: writes a failing test first, then implements the fix

To set this up, just tell Korey what you want the command to do in natural language, and ask it to remember it. The next time you use that keyword, the full sequence fires.

Teams have used this pattern to build commands for sprint kick-off, release prep, bug triage, and more.

The result

A Korey that writes the way your team writes, knows your conventions without being reminded, and runs entire workflows from a single word. The setup takes a few minutes. The payoff is every day after that.