docs(plans): add Spec-3 question-gen-v2 implementation plan

9 tasks: type extension → postprocess → families → sampler_v2 →
generator_v2 → gates → run_store → pipeline_v2 → CLI integration.
Includes structured-logging schemas/metrics and Codex review revisions.

research-wiki/metrics/qgen-gate-rejection-rate.md

@@ -0,0 +1,33 @@
+---
+type: metric
+node_id: metric:qgen-gate-rejection-rate
+title: 出题四门总拦截率
+date: 2026-07-12
+---
+
+# 出题四门总拦截率
+
+## 定义
+
+轻量四门（键验证 + 盲答 + 多真 + 泄漏）在首次尝试中拒绝的题目占比。
+
+## 判定方式
+
+- **类型**: 硬性（上限约束）
+- **阈值**: ≤ 60%（超过说明生成器质量过差，prompt 模板需大幅迭代）
+- **计算**: `COUNT(attempt=1 AND final_status='rejected') / COUNT(attempt=1)`
+- **数据源**: `question_gen_items` WHERE `attempt=1`
+
+## 基线
+
+待首次运行后建立。
+
+## SQL
+
+```sql
+SELECT
+    CAST(SUM(CASE WHEN final_status='rejected' THEN 1 ELSE 0 END) AS REAL) /
+    COUNT(*) AS rejection_rate
+FROM question_gen_items
+WHERE run_id=? AND attempt=1;
+```