Learning Mode

Opt-in coaching mode that turns the agent into a teacher while you build a new project.

Learning mode shifts the agent from doing your work to teaching you how to do it. You stay in the driver’s seat; the agent asks questions, suggests directions, or reviews what you wrote. It is opt-in and intended only for new projects.

What it is

A configuration layer on top of the standard prdwise workflow. When enabled, the agent picks a coaching style based on your profile (technical or writing) and adjusts how it responds — questions instead of answers, hints instead of solutions, reviews instead of rewrites.

When to use it

  • You are starting a new project and want to learn the framework or domain as you go.
  • You want the agent to slow you down on shallow prompts and force you to think.
  • You want a reviewer for code or drafts you wrote yourself.

When to skip it

  • You are working in an existing codebase. Learning mode is new projects only — do not enable it on existing work.
  • You need throughput, not coaching.
  • The task is routine and you already know the answer.

How to enable it

See docs/enable-learning-mode.md for setup. The short version: add the learning-mode rules to your project instructions, pick a profile and coaching style, and start a new project.

Coaching styles

Picked per profile:

  • strict-coach — agent asks questions only, never gives answers. Pushes you to derive the solution.
  • guided-practice — default for technical work. Nudges, syntax hints, debugging questions; small answers when you are stuck.
  • reviewer-mode — you write code, the agent reviews it and flags issues.
  • editor — for writing profiles. Agent edits and critiques your drafts.

The business profile does not offer learning mode.

Shallow prompts

When your prompt is too vague, the agent pushes back with a template:

I want to...
It should be done when...

…followed by three suggestions to sharpen the request. Fill in the blanks before continuing.

Context7

Learning mode pairs with Context7 for fetching current library, framework, and API docs when accuracy matters. Treat it as an opt-in companion, not a replacement for understanding what you are reading.

Reference