constitution: ask questions one at a time for guided UX

Author: michellejw

templates/commands/constitution.md

@@ -18,17 +18,35 @@ You **MUST** consider the user input before proceeding (if not empty).
 
 You are updating the project research constitution at `/memory/constitution.md`. This file defines the scientific context, core principles, and standards for the project. Your job is to collect concrete details from the user and fill in the template.
 
+## Interaction Style
+
+**IMPORTANT**: Ask questions ONE AT A TIME. Do not dump all questions at once.
+
+- After asking a question, STOP and wait for the user's response.
+- Once they answer, update `/memory/constitution.md` with that section, then ask the next question.
+- This creates a guided conversation rather than an overwhelming list.
+
+**Exception**: If the user provides substantial context in their initial input (after the command), use it to pre-fill what you can. Then only ask about gaps, still one question at a time.
+
+**Progress saving**: After each answer, write the updated section to the file so progress is saved even if the conversation is interrupted.
+
 Follow this execution flow:
 
 1. Load the existing constitution at `/memory/constitution.md`.
    - Identify placeholder tokens in square brackets (e.g., `[PROJECT_NAME]`, `[DATA_SOURCES]`).
+   - Check what's already filled in vs. what needs input.
+
+2. If user provided context upfront:
+   - Pre-fill as many sections as possible from their input.
+   - Write the partial constitution to the file.
+   - Identify remaining gaps and ask about them ONE AT A TIME.
 
-2. Collect values for each section:
-   - If user input supplies details, use them.
-   - Otherwise, infer from existing repo context (README, existing scripts, data files).
-   - For sections the user hasn't addressed, ask focused questions or mark as `[TODO: description]`.
+3. If user provided no context (or minimal):
+   - Start with the first section (Research Context).
+   - Ask ONE focused question about that section.
+   - Wait for response, update file, then move to next section.
 
-3. Section-specific guidance:
+4. Section-specific guidance:
 
    **Research Context**: Get the big picture - what questions, why this matters, who uses the outputs.
 
@@ -44,19 +62,19 @@ Follow this execution flow:
 
    **Project Notes**: Probe for constraints that might not be obvious - data embargoes, collaborator agreements, deadlines.
 
-4. Draft the completed constitution:
+5. Draft the completed constitution:
    - Replace placeholders with concrete text.
    - Keep HTML comments as guidance unless the user asks to remove them.
    - Use clear, direct language - this is a working document, not a legal contract.
 
-5. Consistency check:
+6. Consistency check:
    - If `/memory/spec.md` exists, verify data sources and scope align.
    - If analysis scripts exist, verify stated environment matches actual dependencies.
    - Flag any inconsistencies to the user.
 
-6. Write the completed constitution to `/memory/constitution.md`.
+7. Write the completed constitution to `/memory/constitution.md`.
 
-7. Summarize for the user:
+8. Summarize for the user:
    - What sections are complete vs. marked TODO.
    - Any inconsistencies found.
    - Suggested next steps (e.g., "Run `speckit specify` to create your first analysis spec").