docs(questions): v6 — Patch Mode Required, Mass Rewrite Exception

seed/questions/QUESTION_REWRITE_PLAN.md

@@ -162,6 +162,52 @@ Do not write or rewrite 500 daily questions in one pass.
 
 Use a bounded loop. The goal is quality control, not infinite polishing until everyone forgets why the app exists.
 
+### Patch Mode Required
+
+Daily updates must default to patch mode.
+
+Patch mode means:
+
+1. Review the full pack.
+2. Mark only failing question IDs.
+3. Fix only the marked IDs.
+4. Preserve passing questions exactly.
+5. Preserve metadata unless metadata is the failure.
+6. Review the patched questions again.
+7. Repeat only for IDs that still fail.
+
+Do not rewrite all 500 because some questions failed.
+
+Do not rewrite a whole weekday unless the mass rewrite exception applies.
+
+### Mass Rewrite Exception
+
+A mass rewrite is allowed only when more than 60 percent of a weekday or pack fails for the same root cause.
+
+The review report must include:
+
+* failure percentage
+* shared root cause
+* why patching individual IDs is worse
+* fields that will be preserved
+* sample-gate results after the rewrite
+
+Without this report, mass rewriting is not allowed.
+
+### Fix Scope
+
+Every marked question must include one fix scope:
+
+* `prompt_only`
+* `options_only`
+* `prompt_and_options`
+* `metadata_only`
+* `delete_or_replace`
+
+Normal tone fixes should usually be `prompt_only`, `options_only`, or `prompt_and_options`.
+
+Do not change access, depth, tags, or IDs unless the fix scope says metadata is involved.
+
 ### Per Weekday Loop
 
 For each weekday theme:
@@ -209,6 +255,9 @@ When marking a daily question, use one or more of these reasons:
 * weird_option
 * weak_weekday_fit
 * filler_question
+* too_random
+* mechanic_overuse
+* patch_scope_violation
 
 ### Final Pack Gate
 
@@ -222,7 +271,9 @@ After all weekdays are drafted:
 6. Mark anything weak.
 7. Fix marked questions.
 8. Run a second random sample from each weekday.
-9. Ship only if the second sample passes cleanly.
+9. Fix only marked sample items.
+10. Run another sample if anything changed.
+11. Ship only if the second sample passes cleanly and remaining hard flags are 0.
 
 Do not skip the final sample gate.
 
@@ -249,5 +300,7 @@ Each rewrite should provide:
 * updated JSON
 * validation report
 * marked fixes CSV or JSON
+* patched fixes CSV or JSON
+* remaining flags CSV or JSON
 * short review findings
 * apply instructions