Best for
- Solo founders running a product with AI doing the building
- Anyone tired of re-explaining context at the start of every session
- Teams that want one place where work is decided
What you give it
- A goal in plain words — a feature, a fix, a launch step
- Your decisions when it asks (it never invents them)
What you get back
- The goal broken into numbered work items with a status each
- Delegation to the right specialist agent or skill, in order
- A report per item: what changed, how it was tested, what is next
How it works
- Restates the goal and lists what it needs to know before starting.
- Splits the goal into work items small enough to verify one by one.
- Hands each item to the specialist agent or skill that owns that kind of work.
- Collects results, runs the checks, and reports honestly — including what failed.
- Stops and asks you whenever a decision changes money, data or scope.
Example
You: Goal: people should be able to cancel a booking themselves and get the refund the policy allows.
Result: Four work items (rules, service, screen, tests) with owners and order; two questions back — the refund tiers and the cut-off time — then delivery item by item.
Limits — please read
- It coordinates — the quality of the specialists it delegates to still matters.
- It will not invent business rules; unanswered questions pause the item, not the truth.
- It works within one session's context; very long programmes need its written work-item list to carry over.