Clean fragment residue and surface functional audit details

benchmarks/jyotish/reports/drafts/raman_dasha_oracle_draft_2026_06_28.md

@@ -0,0 +1,32 @@
+# Raman Dasha Oracle Draft - 2026-06-28
+
+This draft preserves an untracked pytest-style scratch file that was found
+during workspace residue cleanup. It is not an executable benchmark yet.
+
+## Source Claim
+
+- Source: B. V. Raman, *Hindu Predictive Astrology*.
+- Birth data noted in the scratch file: 1912-08-08, 19:43 IST, Bangalore.
+- Ayanamsa noted in the scratch file: Raman, approximately 21 deg 10 min 23 sec.
+
+## Claimed Vimshottari Boundaries
+
+| Boundary | Target date |
+|----------|-------------|
+| Mars MD start | 1912-08-08 |
+| Rahu MD start | 1918-09-21 |
+| Jupiter MD start | 1936-09-21 |
+| Saturn MD start | 1952-09-21 |
+| Mercury MD start | 1971-09-21 |
+| Ketu MD start | 1988-09-21 |
+
+## Promotion Boundary
+
+Do not move this into `tests/` until:
+
+- the exact book passage or external artifact is cited;
+- Raman ayanamsa mode and dasha-year policy are frozen;
+- the engine call is real, not a mock;
+- the tolerance is explicit and reviewed.
+
+Until then it is a benchmark/oracle draft only.

docs/benchmark/legacy-marriage-v6.1/README.md

@@ -0,0 +1,14 @@
+# Legacy Marriage Benchmark v6.1
+
+This folder preserves the May 2026 marriage validation artifacts that were
+previously left under `tests/` as untracked residue.
+
+These files are research and benchmark evidence, not executable pytest tests:
+
+- `verify-results-v6.1.json` contains the 18-case structured validation output.
+- `印度占星实战案例综合验证报告-v6.1-2026-05-03.md` contains the corresponding Rao
+  8-parameter marriage validation report.
+
+Keep this folder as a historical benchmark source. Promote individual rules into
+tracked tests only after the target oracle fields, input data, and acceptance
+tolerances are explicitly frozen.

docs/benchmark/legacy-marriage-v6.1/印度占星实战案例综合验证报告-v6.1-2026-05-03.md

@@ -0,0 +1,139 @@
+# 印度占星 Skill 实战案例综合验证报告 v6.1 (Rao 8参数+多维度)
+
+> **日期**: 2026-05-03 | **案例**: 18名人 | **核心**: Rao 8参数体系 + 多维度交叉验证
+> **v6.0→v6.1**: 补全18项遗漏（Rao P1-P8, UL/Argala/D7/D60/Vivah Saham等）
+
+## 一、Rao 8参数命中率统计
+
+**总婚姻事件**: 26
+
+| 参数 | 描述 | 命中数 | 命中率 | Rao原始命中率 |
+|------|------|--------|--------|-------------|
+| P1 | Vimshottari PAC连接 | 25 | 25/26=96% | 100% |
+| P2 | Chara Dasha+Jaimini | 19 | 19/26=73% | 96% |
+| P3 | Vivah Saham | 15 | 15/26=58% | 77% |
+| P4 | Double Transit PAC | 17 | 17/26=65% | 85% |
+| P5 | Transit LL-7L连接 | 17 | 17/26=65% | 98% |
+| P6 | Jupiter激活性别星 | 18 | 18/26=69% | 68% |
+| P7 | 行星聚集Lagna/7H | 4 | 4/26=15% | 70% |
+| P8 | Transit LL/7L互换 | 4 | 4/26=15% | 59% |
+
+**累积命中分布**:
+- 6+参数: 5/26 = 19.2% (Rao: 85.78%)
+- 5+参数: 13/26 = 50.0% (Rao: 94.50%)
+- 4+参数: 22/26 = 84.6% (Rao: 100%)
+
+## 二、逐案例 Rao 8参数得分
+
+| 名人 | 配偶 | 日期 | P1 | P2 | P3 | P4 | P5 | P6 | P7 | P8 | 总计 |
+|------|------|------|----|----|----|----|----|----|----|----|------|
+| Tom Cruise | Katie Holmes |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✓ | **7/8** |
+| Elon Musk | Justine Wilson |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✓ | ✓ | **7/8** |
+| Britney Spears | Kevin Federline |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✓ | **7/8** |
+| Albert Einstein | Mileva Maric |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ | **6/8** |
+| Nelson Mandela | Winnie Madikizela |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ | **6/8** |
+| Albert Einstein | Elsa Einstein |  | ✓ | ✓ | ✗ | ✗ | ✓ | ✓ | ✗ | ✓ | **5/8** |
+| Princess Diana | Prince Charles |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | **5/8** |
+| Michael Jackson | Debbie Rowe |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | **5/8** |
+| Tom Cruise | Mimi Rogers |  | ✓ | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | ✗ | **5/8** |
+| Tom Cruise | Nicole Kidman |  | ✓ | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | ✗ | **5/8** |
+| Elon Musk | Talulah Riley |  | ✓ | ✓ | ✗ | ✓ | ✗ | ✓ | ✓ | ✗ | **5/8** |
+| Shah Rukh Khan | Gauri Khan |  | ✓ | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | ✗ | **5/8** |
+| Amitabh Bachchan | Jaya Bhaduri |  | ✓ | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | ✗ | **5/8** |
+| Steve Jobs | Laurene Powell |  | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✓ | ✗ | **4/8** |
+| Barack Obama | Michelle Robinson |  | ✓ | ✗ | ✗ | ✓ | ✗ | ✓ | ✓ | ✗ | **4/8** |
+| Bill Gates | Melinda French |  | ✓ | ✗ | ✓ | ✗ | ✓ | ✓ | ✗ | ✗ | **4/8** |
+| Nelson Mandela | Evelyn Mase |  | ✓ | ✓ | ✗ | ✗ | ✓ | ✓ | ✗ | ✗ | **4/8** |
+| Oprah Winfrey | Stedman Graham |  | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | **4/8** |
+| Jeff Bezos | MacKenzie Scott |  | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | **4/8** |
+| Britney Spears | Jason Alexander |  | ✓ | ✓ | ✓ | ✗ | ✗ | ✓ | ✗ | ✗ | **4/8** |
+| Mark Zuckerberg | Priscilla Chan |  | ✓ | ✓ | ✗ | ✓ | ✗ | ✓ | ✗ | ✗ | **4/8** |
+| Narendra Modi | Jashodaben Modi |  | ✓ | ✓ | ✗ | ✗ | ✓ | ✓ | ✗ | ✗ | **4/8** |
+| Michael Jackson | Lisa Marie Presley |  | ✓ | ✗ | ✓ | ✗ | ✗ | ✓ | ✗ | ✗ | **3/8** |
+| Nelson Mandela | Graca Machel |  | ✓ | ✗ | ✓ | ✗ | ✗ | ✓ | ✗ | ✗ | **3/8** |
+| Prince Harry | Meghan Markle |  | ✗ | ✗ | ✓ | ✓ | ✗ | ✗ | ✗ | ✗ | **2/8** |
+| Priyanka Chopra | Nick Jonas |  | ✓ | ✗ | ✗ | ✗ | ✓ | ✗ | ✗ | ✗ | **2/8** |
+
+## 三、多维度交叉验证摘要
+
+### 3.1 UL (Upapada Lagna) 婚姻稳定性
+
+| 名人 | UL | UL主 | UL第2宫 | 吉星 | 凶星 | 稳定性 | DK-UL关系 |
+|------|-----|------|---------|------|------|--------|----------|
+| Albert Einstein | Pisces | Jupiter | Aries | - | - | neutral | UL=DK sign |
+| Steve Jobs | Pisces | Jupiter | Aries | - | Mars | weak | no direct link |
+| Princess Diana | Taurus | Venus | Gemini | Mercury | Sun | neutral | UL=DK sign |
+| Barack Obama | Capricorn | Saturn | Aquarius | - | Ketu | weak | UL=DK sign |
+| Bill Gates | Virgo | Mercury | Libra | Venus | Sun,Saturn | weak | no direct link |
+| Michael Jackson | Cancer | Moon | Leo | Mercury | Sun | neutral | no direct link |
+| Nelson Mandela | Virgo | Mercury | Libra | Moon | - | strong | no direct link |
+| Tom Cruise | Taurus | Venus | Gemini | - | Sun | weak | UL=DK sign |
+| Elon Musk | Taurus | Venus | Gemini | Mercury | Sun | neutral | no direct link |
+| Oprah Winfrey | Scorpio | Mars | Sagittarius | - | - | neutral | UL=DK sign |
+| Prince Harry | Scorpio | Mars | Sagittarius | Jupiter | - | strong | no direct link |
+| Jeff Bezos | Aquarius | Saturn | Pisces | Jupiter | - | strong | no direct link |
+| Priyanka Chopra | Gemini | Mercury | Cancer | - | Sun | weak | no direct link |
+| Shah Rukh Khan | Capricorn | Saturn | Aquarius | - | Saturn | weak | no direct link |
+| Britney Spears | Virgo | Mercury | Libra | Jupiter | - | strong | no direct link |
+| Mark Zuckerberg | Taurus | Venus | Gemini | - | - | neutral | UL=DK sign |
+| Narendra Modi | Virgo | Mercury | Libra | - | - | neutral | UL=DK sign |
+| Amitabh Bachchan | Taurus | Venus | Gemini | - | - | neutral | no direct link |
+
+### 3.2 Argala 7宫净评估
+
+| 名人 | 支持(Argala) | 阻碍(Virodha) | 净分 | 评估 |
+|------|------------|-------------|------|------|
+| Albert Einstein | Ketu(H2) | Sun(H10),Mercury(H10),Venus(H10),Saturn(H10) | -3 | block |
+| Steve Jobs | Jupiter(H11),Ketu(H11) | Saturn(H3) | 1 | support |
+| Princess Diana | Moon(H4),Ketu(H4) | Mars(H10),Jupiter(H3),Saturn(H3),Rahu(H10) | -2 | block |
+| Barack Obama | Ketu(H2) | - | 1 | support |
+| Bill Gates | Sun(H4),Jupiter(H2),Venus(H4),Saturn(H4),Ketu(H11) | Mars(H3),Mercury(H3) | 3 | support |
+| Michael Jackson | Mars(H11),Venus(H2),Ketu(H11) | Sun(H3),Mercury(H3) | 1 | support |
+| Nelson Mandela | Jupiter(H2) | Sun(H3),Mercury(H3),Saturn(H3) | -2 | block |
+| Tom Cruise | Saturn(H4),Ketu(H4) | Moon(H10),Venus(H10),Rahu(H10) | -1 | block |
+| Elon Musk | Ketu(H2) | Moon(H3),Venus(H12),Saturn(H12) | -2 | block |
+| Oprah Winfrey | Sun(H2),Mercury(H2),Venus(H2),Saturn(H11),Rahu(H2) | Moon(H12),Mars(H12) | 3 | support |
+| Prince Harry | Saturn(H11) | Mars(H12),Venus(H10),Ketu(H12) | -2 | block |
+| Jeff Bezos | Sun(H2),Moon(H2),Mercury(H2),Venus(H4),Ketu(H2) | Mars(H3),Saturn(H3) | 3 | support |
+| Priyanka Chopra | Mars(H11),Saturn(H11),Ketu(H2) | Jupiter(H12) | 2 | support |
+| Shah Rukh Khan | Mars(H4),Mercury(H4),Jupiter(H11),Ketu(H4) | Sun(H3),Rahu(H10) | 2 | support |
+| Britney Spears | Moon(H11),Venus(H11),Ketu(H11) | - | 3 | support |
+| Mark Zuckerberg | Moon(H2),Mars(H2),Jupiter(H4),Saturn(H2) | Ketu(H3) | 3 | support |
+| Narendra Modi | Moon(H2),Mars(H2),Venus(H11),Saturn(H11) | Sun(H12),Mercury(H12),Ketu(H12) | 1 | support |
+| Amitabh Bachchan | Saturn(H4) | - | 1 | support |
+
+### 3.3 Vipareeta Raja Yoga 检测
+
+- **Albert Einstein**: 6L(Mars)在H8 → Vipareeta
+- **Steve Jobs**: 12L在8宫: 死亡征象
+- **Michael Jackson**: 8L(Saturn)在H6 → Vipareeta
+- **Tom Cruise**: 12L在8宫: 死亡征象
+- **Elon Musk**: 6L(Mars)在H8 → Vipareeta, 8L(Saturn)在H12 → Vipareeta
+- **Oprah Winfrey**: 8L(Moon)在H12 → Vipareeta
+- **Priyanka Chopra**: 12L在8宫: 死亡征象
+
+## 四、置信度评估 [A/B/C]
+
+### [A] 已验证
+1. DK Nakshatra → 配偶深层特质: 100%
+2. D9 DK → 配偶本质属性: 100%
+3. Vimshottari Dasha计算: 100%
+4. Rao P1 (Vimshottari PAC): 100% (218案例)
+5. Rao P5 (Transit LL-7L): 98% (214/218)
+
+### [B] 可验证
+1. Rao P2 (Chara Dasha): 96%
+2. Rao P4 (Double Transit PAC): 85%
+3. Rao P3 (Vivah Saham): 77%
+4. 7宫主落宫 → 相遇方式: ~75%
+5. Yogakaraka → 事业高峰
+6. UL第2宫 → 婚姻稳定性
+
+### [C] 待验证
+1. Rao P6/P7/P8: 59-70%
+2. Vipareeta Raja Yoga: 少案例
+3. 12宫主落8宫死亡征象: 少案例
+
+---
+*报告生成: 2026-05-03 | 验证脚本: verify-jyotish-v6.1.py*
+*数据源: Swiss Ephemeris (Lahiri Ayanamsa) | 18名人案例*

findings.md

@@ -10,6 +10,19 @@
 - Round 25 副手任务已发布，扫描时已看到部分 `antigravity_round25_*` 报告开始生成，但未满 18+ 前不能视为完成。
 - Ashtakoot 结论边界：Round 24 “全 0/瞎编”属于需纠正的过强说法；当前 `scripts/ashtakoot.py` 有非零本地规则，但外部 oracle 仍 0/5，不能声称与 JHora/AstroSage/VedAstro 完全一致。
 
+## 2026-06-28 全仓工作流遗漏扫描结论
+
+- 当前排除 venv/build/dist/cache 后仍有约 1528 个文件；主风险区是 `docs/`、`references/`、`scripts/`、`tests/`、`jyotish-app/` 与 `scratch/`。
+- `python3 scripts/audit_fragments.py --strict` 当前返回 `valid: true`、`problem_count: 0`，注册表 89 技法均有引擎/API/前端/测试/脚本中的至少一种产品表面；这说明“registry 声称但完全无入口”的大类问题暂未发现。
+- 扫描发现真实遗漏：Functional Benefic/Malefic 后端和 MCP 已接入，但前端 `Technique Audit Table`/Skill Map 一度缺少可见审计行；已通过 `tests/test_frontend_productization.py::test_skill_map_surfaces_functional_benefic_malefic_audit_row` 守门。
+- `audit_fragments` 仍标记 3 个脚本候选碎片：`oracle_functional_benefics.py`、`patch_api_tz.py`、`patch_engine_tz.py`。其中 `oracle_functional_benefics.py` 是功能性吉凶 CLI 包装器，应决定是否纳入 registry/quality gate；`patch_api_tz.py` 与 `patch_engine_tz.py` 是会改源码的一次性补丁脚本，不应作为产品工作流留在 `scripts/` 默认面。
+- 当前 Git 未跟踪残留 11 个：个人/临时输出 `full_chart_data.json`、`test_dasha.json`、`test_output.json`、`scratch_extract.py`、`scratch_mcp_eval.py`；一次性补丁 `scripts/patch_api_tz.py`、`scripts/patch_engine_tz.py`；个人同步工具 `scripts/sync_to_workbuddy.sh`；测试候选 `tests/test_dasha_raman_truth.py`；测试/研究 artifact `tests/verify-results-v6.1.json` 与 `tests/印度占星实战案例综合验证报告-v6.1-2026-05-03.md`。
+- `docs/research/ACTIVE_FRONTS.md` 当前仍列出未闭合项：Vimsopaka semantic mapping for `NEECHA_BHANGA / GREAT_FRIEND / GREAT_ENEMY`，以及 functional role 的 Technique Audit Table rendering 跟进。
+- `docs/research/vedastro_parity_matrix_latest.md` 当前 13 行中 `partial=8`、`covered=4`、`missing=1`。P0 未闭合集中在 Tajika Annual、Ayanamsa parity、Report Rendering、MCP/API VedAstro live adapter smoke、Ashtakavarga/Shadbala parity、EventsAtRange/Life Event Graph；Numerology/Non-Jyotish Tools 为 P2 adjacent missing，不属于 Jyotish 主工作流。
+- 验证命令：`python3 scripts/run_quality_gate.py --profile quick --skip-frontend-runtime` 通过；`npm run build` 通过；`python3 -m pytest tests/ -q` 通过；`git diff --check` 通过。
+- 后续收口结论：`oracle_functional_benefics.py` 已通过 CLI JSON 合同测试进入正式测试表面；`patch_api_tz.py`、`patch_engine_tz.py` 与个人 scratch/output 已归入 ignored `scratch/local`；v6.1 婚恋验证资产已转入 `docs/benchmark/legacy-marriage-v6.1/`；Raman Dasha 草稿保留为 benchmark draft，不再伪装成 pytest。
+- 新发现的真实遗漏：relationship strict bridge 原先只识别字符串 `"BadConstellations": "good"`，不识别 nested dict 形态，也不读取 `exceptions` mitigation 文本；已补 `exception_mitigated_match` 与 nested Kuta 归一化，避免 Synastry/Ashtakoot 资产被浅层解析吞掉。
+
 ## 开源与本地基线
 
 - `references/open-source-jyotish-scan-2026.md` 已记录可直接复用/对标项目：dashaflow、jyotishganit、panchanga_api、jaimini-tropical、VedicAstro、KPAstroDashboard、vedic_astro_npm、PyJHora、VedAstro、xalen-ephemeris。

jyotish-app/skill-map.js

@@ -37,7 +37,6 @@ const TECHNIQUES = [
   ['Functional Benefic/Malefic / 功能吉凶星', '已接入', '按 Lagna 输出功能吉星、功能凶星、Yogakaraka 与中性星，并进入 Technique Audit Table'],
   ['Argala', '已接入', '综合深度 Tab 已展示 Argala/Virodhargala 支持、阻挡和净分'],
   ['Shadbala', '已接入', '六维力量表；保留绝对值校准边界'],
-  ['Functional Benefic/Malefic', '已接入', '功能吉凶星判定已进入高严谨工作流，并在 Technique Audit Table 中影响结论置信度'],
   ['Ashtakavarga', '已接入', 'SAV/BAV、Transit AV'],
   ['KP / Sub-lord', '已接入', 'KP页已支持问题域筛选、重点宫位 ABCD significator、高权重证据解释、Sublord 表格与使用边界'],
   ['Synastry / Ashtakoot', '已接入', '合盘页已支持对方完整出生资料排盘、36分Ashtakoot、D9专题、Kuja Dosha平衡与Dasha同步；保留月亮黄经快算'],

mcp_server.py

@@ -494,10 +494,21 @@ def _derive_synastry_relationship_support(modules: Dict[str, Any]) -> Dict[str,
         signals.append("ashtakoot_approved")
     if total_score is not None and total_score >= 27:
         signals.append("ashtakoot_high_score")
+    exceptions = synastry.get("exceptions")
+    if isinstance(exceptions, list):
+        exception_text = " ".join(str(item).lower() for item in exceptions)
+        if "mitigat" in exception_text or "exception" in exception_text:
+            signals.append("exception_mitigated_match")
     additional_kutas = synastry.get("additional_kutas")
     if isinstance(additional_kutas, dict):
-        vedha_good = additional_kutas.get("Vedha") == "good"
-        bad_constellations_good = additional_kutas.get("BadConstellations") == "good"
+        def _kuta_result(value: Any) -> str | None:
+            if isinstance(value, dict):
+                result = value.get("result")
+                return str(result).lower() if result is not None else None
+            return str(value).lower() if value is not None else None
+
+        vedha_good = _kuta_result(additional_kutas.get("Vedha")) == "good"
+        bad_constellations_good = _kuta_result(additional_kutas.get("BadConstellations")) == "good"
         rajju = additional_kutas.get("Rajju")
         rajju_good = isinstance(rajju, dict) and rajju.get("result") == "good"
         if vedha_good and bad_constellations_good and rajju_good:

progress.md

@@ -405,3 +405,21 @@
 - 更新 `docs/research/ACTIVE_FRONTS.md` 的 Oracle Closure 入口，要求后续修改 oracle-dependent adjudicator 或 benchmark claims 前先运行 `python3 scripts/oracle_benchmark_inventory.py --format json`。
 - 收口上一轮遗留的 finance Shadbala 组件审计：`mcp_server.py` 对 `shadbala.planets` 要求 `sthana/dig/kala/chesta/naisargika/drik` 六分量，缺失时 `confidence_cap = low` 且进入 `secondary_context += ["shadbala_component_gap"]`；补文档 `docs/research/shadbala_component_confidence_cap_v1_2026_06_28.md`。
 - 验证：`python3 -m pytest tests/test_oracle_benchmark_inventory.py -q` 通过；`python3 -m pytest tests/test_mcp_strict_workflow_finance.py -q` 24 项通过；`python3 -m json.tool docs/research/oracle_benchmark_inventory_latest.json` 通过。
+## 2026-06-28T18:27:42+08:00 - 全仓工作流遗漏扫描
+
+- 读取并恢复 `task_plan.md`、`findings.md`、`progress.md`，确认当前已有“整机与 Git 云端地毯式遗漏审计”上下文。
+- 执行仓库规模扫描：排除 venv/build/dist/cache 后约 1528 个文件，重点检查 docs/references/scripts/tests/jyotish-app/scratch。
+- 执行 `python3 scripts/audit_fragments.py --strict`：注册表 89 技法的产品表面审计通过，问题数 0；候选碎片仍为 `oracle_functional_benefics.py`、`patch_api_tz.py`、`patch_engine_tz.py`。
+- 执行 quick quality gate 时发现 `test_skill_map_surfaces_functional_benefic_malefic_audit_row` 红灯，根因为功能性吉凶层未在前端 Skill Map / Technique Audit Table 可见化；补齐后聚焦测试通过。
+- 分类未跟踪残留 11 个，其中多数为 scratch/个人输出；`tests/test_dasha_raman_truth.py` 是未纳入 CI 的测试候选；`tests/verify-results-v6.1.json` 和中文报告应移出 tests 或明确接入测试。
+- 复核 `ACTIVE_FRONTS.md` 与 `vedastro_parity_matrix_latest.md`：仍有 Vimsopaka 高阶语义映射、Functional role 审计表渲染后续、VedAstro adapter endpoint smoke、Life Event Graph、报告渲染等未闭合工作。
+- 验证：`python3 scripts/run_quality_gate.py --profile quick --skip-frontend-runtime` 通过；`npm run build` 通过；`python3 -m pytest tests/ -q` 通过；`git diff --check` 通过。
+
+## 2026-06-28T18:58:00+08:00 - 碎片残留与 Functional/Synastry 桥接收口
+
+- 将散落在根目录和 `scripts/` 的临时输出、scratch 脚本、一次性 timezone patch 脚本移动到 ignored `scratch/local/...`，不再污染产品脚本面。
+- 将旧 `tests/verify-results-v6.1.json` 与中文婚恋验证报告移动到 `docs/benchmark/legacy-marriage-v6.1/`，新增 README 标明它们是历史 benchmark evidence，不是 pytest。
+- 将未成熟的 Raman Dasha oracle pytest 草稿改写为 `benchmarks/jyotish/reports/drafts/raman_dasha_oracle_draft_2026_06_28.md`，保留目标日期但明确 promotion boundary，避免被误纳入 CI。
+- `oracle_functional_benefics.py` 新增 CLI JSON 合同测试；`python3 scripts/audit_fragments.py --strict` 显示 candidate_count 从 3 降到 0。
+- Functional Benefic/Malefic 的 `Technique Audit Table` 与 real-reading checklist 现在显式要求/输出 `functional_neutrals` 与 `yogakarakas`。
+- 合婚 strict bridge 修复 `additional_kutas.BadConstellations` 的 nested dict 形态，并把 mitigation exceptions 折叠成 `exception_mitigated_match`，防止已有 Ashtakoot 细节资产在 relationship adjudicator 中被吞掉。

references/ai-reading-workflow-prompt.md

@@ -428,6 +428,15 @@ AI不应该凭空假设一个时间，而应该通过结构化互动帮助用户
 
 在使用下方意图路由表之前，先读取 `strict-workflow-router.md`。凡涉及事业、婚恋、财务、事件应期、历史回测、技法可靠性或高级分析，必须选择对应 strict route，并在最终输出附 Technique Audit Table。
 
+在 Technique Audit Table 中，`Functional Benefic/Malefic` 不得只写“已使用”。必须显式列出：
+- functional benefics
+- functional malefics
+- functional neutrals
+- yogakarakas
+- 对结论置信度的影响
+
+若缺少其中任一层，不得伪装成高严谨完成态，应降级置信度或标记 `blocked`。
+
 | 用户意图 | Strict route | 必须额外完成 |
 |----------|-------------|-------------|
 | 事业机会/职业方向/项目落地 | `career-timing-strict` | D1+D9+D10+Vimshottari+Jaimini+AmK/Karakamsha+AL/A10+Shadbala+AV+Argala |

references/real-reading-quality-checklist.md

@@ -26,33 +26,34 @@
 
 11. 是否显式给出 `Functional Benefic/Malefic`？
 12. 是否区分了自然属性、功能属性、题目域角色、当前激活结果？
-13. 对 `Saturn / Mars / Rahu / Ketu` 是否避免了“天然负面”式偷懒断法？
+13. 是否显式列出 `functional neutrals` 与 `Yogakaraka`，避免它们在最终报告中被折叠丢失？
+14. 对 `Saturn / Mars / Rahu / Ketu` 是否避免了“天然负面”式偷懒断法？
 
 ## E. 分盘层
 
-14. 婚恋是否至少展开 `D9 / UL / DK`？
-15. 事业是否至少展开 `D10 / A10 / AmK/Karakamsha`？
-16. 财富是否至少展开 `D2/D11`，并回看 D1/D9/D10 的变现链？
-17. 若使用 `D30 / D60`，是否明确说明这是高敏后置层而非主裁判？
+15. 婚恋是否至少展开 `D9 / UL / DK`？
+16. 事业是否至少展开 `D10 / A10 / AmK/Karakamsha`？
+17. 财富是否至少展开 `D2/D11`，并回看 D1/D9/D10 的变现链？
+18. 若使用 `D30 / D60`，是否明确说明这是高敏后置层而非主裁判？
 
 ## F. Timing 层
 
-18. 是否至少执行 `Vimshottari + Narayana` 双轨交叉？
-19. 若做具体事件判断，是否加入 `Transit / Double Transit / annual / KP or Chara support`？
-20. 如果 timing 只做到窗口级，而不是日期级，是否说清了边界？
+19. 是否至少执行 `Vimshottari + Narayana` 双轨交叉？
+20. 若做具体事件判断，是否加入 `Transit / Double Transit / annual / KP or Chara support`？
+21. 如果 timing 只做到窗口级，而不是日期级，是否说清了边界？
 
 ## G. 外部真值与诚信边界
 
-21. 是否清楚声明 `production_tuning_allowed` 的当前状态？
-22. 是否把外部 oracle 未闭环的前线（Dasha boundary / Shadbala absolute / Tajika annual）说清楚？
-23. 是否避免把内部一致性伪装成传统软件级真值冻结？
+22. 是否清楚声明 `production_tuning_allowed` 的当前状态？
+23. 是否把外部 oracle 未闭环的前线（Dasha boundary / Shadbala absolute / Tajika annual）说清楚？
+24. 是否避免把内部一致性伪装成传统软件级真值冻结？
 
 ## H. 输出层
 
-24. 是否输出 `Technique Audit Table` 或等价审计表？
-25. 是否给出 `used / partial / blocked / not used` 级别说明？
-26. 是否给出置信度，而不是只给断语？
-27. 是否把 must-not-overclaim 边界说清楚？
+25. 是否输出 `Technique Audit Table` 或等价审计表？
+26. 是否给出 `used / partial / blocked / not used` 级别说明？
+27. 是否给出置信度，而不是只给断语？
+28. 是否把 must-not-overclaim 边界说清楚？
 
 ## 一句话用法
 

scripts/jyotish_engine.py

@@ -893,6 +893,8 @@ def _build_technique_audit_table(functional_layer, oracle_progress, modules):
             'note': (
                 f"关键功能吉星={functional_layer.get('functional_benefics', [])}; "
                 f"关键功能凶星={functional_layer.get('functional_malefics', [])}; "
+                f"功能中性星={functional_layer.get('functional_neutrals', [])}; "
+                f"Yogakaraka={functional_layer.get('yogakarakas', [])}; "
                 f"{functional_layer.get('effect_on_confidence', '高严谨模式缺少功能性吉凶星判定。')}"
             ),
         },

tests/test_cli_smoke.py

@@ -285,6 +285,8 @@ def test_full_reading_reports_ayanamsa_metadata_and_ai_prompt_pack() -> None:
     assert "高严谨" in functional_rows[0]["note"]
     assert "关键功能吉星=" in functional_rows[0]["note"]
     assert "关键功能凶星=" in functional_rows[0]["note"]
+    assert "功能中性星=" in functional_rows[0]["note"]
+    assert "Yogakaraka=" in functional_rows[0]["note"]
     dasha_rows = [row for row in audit_table if row["technique"] == "Vimshottari + Narayana Cross-check"]
     assert dasha_rows
     assert dasha_rows[0]["status"] == "used"

tests/test_mcp_strict_workflow_functional_layer.py

@@ -3,6 +3,10 @@
 
 from __future__ import annotations
 
+import json
+import subprocess
+import sys
+
 from mcp_server import _collect_strict_evidence
 from scripts.functional_benefics import derive_functional_benefic_malefic
 
@@ -74,3 +78,27 @@ def test_relationship_functional_layer_uses_ascendant_even_when_planets_missing(
     assert "Sun" in functional["functional_benefics"]
     assert "Venus" in functional["functional_malefics"]
     assert "functional_benefic_malefic_used" in strict["event_judgement"]["secondary_context"]
+
+
+def test_oracle_functional_benefics_cli_exposes_json_contract() -> None:
+    completed = subprocess.run(
+        [
+            sys.executable,
+            "scripts/oracle_functional_benefics.py",
+            "--ascendant",
+            "Leo",
+            "--format",
+            "json",
+        ],
+        check=True,
+        capture_output=True,
+        text=True,
+    )
+    report = json.loads(completed.stdout)
+
+    assert report["status"] == "used"
+    assert report["ascendant"] == "Leo"
+    assert "Sun" in report["functional_benefics"]
+    assert "Venus" in report["functional_malefics"]
+    assert "Mars" in report["yogakarakas"]
+    assert report["source"] == "strict_functional_benefic_malefic_v1"

tests/test_mcp_strict_workflow_relationship.py

@@ -167,6 +167,34 @@ def test_relationship_synastry_bridge_cannot_lift_label_when_core_marriage_layer
     assert "upapada_lagna" in strict["missing_evidence"]
 
 
+def test_relationship_synastry_bridge_understands_nested_bad_constellations_and_exception_mitigation() -> None:
+    result = _base_relationship_result()
+    result["modules"]["synastry"] = {
+        "total_score": 18.0,
+        "max_score": 36.0,
+        "is_approved": True,
+        "exceptions": ["Nadi Dosha mitigated by good Bhakoot and Rajju."],
+        "additional_kutas": {
+            "Mahendra": "good",
+            "StreeDeergha": "good",
+            "Vedha": "good",
+            "Rajju": {"result": "good", "group": None, "effect": ""},
+            "BadConstellations": {"result": "good", "issues": []},
+        },
+    }
+
+    strict = _collect_strict_evidence("relationship", result)
+
+    assert strict["present_evidence"]["synastry_relationship_support"] == {
+        "level": "moderate",
+        "source": "synastry_relationship_bridge_v1",
+        "signals": ["ashtakoot_approved", "exception_mitigated_match", "kuta_exception_clean"],
+        "total_score": 18.0,
+        "approved": True,
+    }
+    assert "synastry_support" in strict["event_judgement"]["secondary_context"]
+
+
 def test_relationship_dignity_guardrail_ignores_non_relevant_planets() -> None:
     result = _base_relationship_result()
     result["modules"]["chart"] = {

tests/test_real_reading_quality_checklist.py

@@ -20,6 +20,8 @@ def test_real_reading_quality_checklist_exists_and_covers_structural_gaps() -> N
         "covered",
         "complete",
         "Functional Benefic/Malefic",
+        "Yogakaraka",
+        "functional neutrals",
         "Vimshottari + Narayana",
         "production_tuning_allowed",
         "Technique Audit Table",