stage-planlisted

# Plan one story Input: the story file, the project's product scope (`backlog/product.md`, or where the profile's `product:` points), and its glossary and context map if it has them. Nothing else — not the chat history, not an earlier run. Output: `tasks/<story>/plan.md`, and nothing else. You write no test and no production code. ## Do 1. Read the story: its context, its acceptance criteria, its open assumptions. 2. Locate that bounded context in the code. If the project keeps a context map and the story's `context` is not in it, stop: write the plan file with a `## needs-human` section stating that the story needs a new context or a new relationship between contexts. That is a scoping decision. A project without a context map is not blocked by this — note the absence and go on. 3. **Take the product decisions from the product scope.** Its `## Surfaces`, `## How it works` and `## Look and feel` answer where state lives, which surfaces the product has and how a page looks. Plan within them. A question they already answer is not a question for the human; one they leave open, or a story that contradicts them, is — through `## needs-human`, like any other. Without a product scope, say so in the plan and plan from the story alone. 4. **Check that the story's actor can already reach the behaviour.** The criteria name someone — a customer, an operator, an administrator — and a way in: a page, an endpoint, a message, a command line. If the context has n