How Spark coaches a learning proposal

On the proposal page, there's a Coach My Idea section. The kid types a rough description of what they want to do — "I want to learn piano" or "can I do something with cooking?" — and optionally adds context in the details field. Then they hit Coach My Idea.

Spark returns a draft that sits next to the kid's current proposal. It includes a title, description, category, reward suggestion, task kind, and — if it's a learning task — subject, goal, level, estimate, proof checklist, and resources. The kid can apply individual fields, apply everything, or discard the draft entirely.

Spark can also revise an existing proposal in progress. On a draft that's already taking shape, the coach reads what's there and suggests improvements rather than starting from scratch.

When you hit Coach My Idea, Spark receives the rough idea and details, the child's learning profile (interests, goals, levels, formats, constraints), the family learning profile (values, blocked topics, parent notes), available categories and the reward range, and a list of recently completed or rejected tasks.

The recent task history is there so Spark doesn't suggest something the kid already finished or the family already turned down. It's a small thing, but it matters.

Spark doesn't remember between proposals. Each coaching request starts fresh. What carries over is the profile — not anything from a prior conversation.

If Spark shapes the proposal as a learning task, it identifies a few search topics based on the subject and goal, then searches for resources filtered by your family's allowed types and blocked domains.

These are filtered before they reach the proposal. Nothing from a blocked domain will appear. Resource types you've turned off won't come back. If nothing survives the filters for a given topic, no resources are returned — Spark won't fill the space with something that didn't pass your rules.

Resources show up with a title, a short description, and where they came from. Visibility — whether the kid can see them or only a parent can — is controlled in the task detail after the proposal is approved.

Spark may flag a few things in a Warnings section: if something similar is already assigned or was recently finished, if the proposal touches a topic you've blocked, or if the idea is too vague to evaluate as a task.

Warnings are visible to the parent when they review the proposal. The kid still decides whether to adjust or submit anyway — warnings don't prevent submission.

Spark doesn't approve anything. Every proposal still goes through the parent queue, same as a regular chore. The coaching is preparation for that review, not a shortcut around it.

Spark won't invent resources it can't find, override blocked domains, or make decisions that belong to the family. It works with the context you've given it — if the profiles are thin, the suggestions will be thinner.