questionnairelisted

# questionnaire Batch the conversation's unresolved decisions into one HTML form; consume the markdown answers it emits. The answers file is the alignment contract — every question comes back as answered, open, or skipped-by-branching; nothing is silently dropped. ## Workflow 1. Mine the discussion, notes, and memory for *decisions still open*, not topics. Points already settled go into `intro` as a "current understanding" summary — add one leading `text` question asking what in it is wrong, instead of re-asking each settled point. 2. Write `questions.json` (schema below), then build: ```bash python3 scripts/build.py questions.json -o alignment.html ``` 3. Deliver: write the HTML somewhere the user can open it (project dir or scratchpad) and give them the path; in a remote/claude.ai session, publish it as an Artifact instead. 4. Consume: the page's **Generate answers** button emits markdown the user pastes back or downloads as `.md`. Parse by the `` (`id`) `` keys, not question numbers or wording. Treat `open — no answer given` as still-undecided — never fill those in yourself. Record ruled decisions wherever the project keeps them (design doc, ADR, memory), quoting the answer. ## questions.json ```json { "title": "Mitate — design alignment", "intro": "Current understanding (flag anything wrong in Q1):\n- ...", "outro": "Closing instruction echoed to the agent under the answers.", "questions": [ { "id": "storage", "label": "Storage layo