ask-then-build

Phase 1 — Questions

1. Identify the 3-6 most non-obvious open questions about the feature or change: edge cases, where it lives in the UI, failure behavior, scope boundaries, how it interacts with existing rules.
2. Ask them ONE at a time. For each: the question, top options A-D, your preferred pick with a one-line reason. Then stop and wait. Use this exact layout. Put a blank line between EVERY block. Markdown renderers (bb included) collapse single newlines into one paragraph, so options written back-to-back render as one run-on blob.
   Question title?
   One line of context, if needed.
   A. First option.
   B. Second option.
   C. Third option.
   D. Fourth option.
   My pick: A. One-line reason.
   Never number questions as "1 of N" unless you truly know N. If the count is open, keep it open and keep asking until you have the context you need.
3. When the user answers, record the decision immediately — update the repo's docs (product requirements, ADR, or README) if the project has them.
4. If the user overrides an earlier documented decision, update the docs right away and say what was superseded.

Phase 2 — Prompt

1. Read-first files: the authoritative docs (AGENTS.md, requirements, ADRs).
2. What to build: numbered implementation steps, concrete file-level where useful.
3. How to validate: build, lint, and a manual check with real data.
4. Rules: don't commit, report back with files changed.

Style

• Very concise, plain English, short sentences.
• Never bundle questions. Never write the prompt before all answers are in.
• Never put two options on one line. Always a blank line between A, B, C, D and before "My pick".
• Keep the prompt to a single paragraph — if it needs two, the scope is too big; say so.