@boyingliu01/xp-gate 0.18.4 → 0.19.0

package/skills/delphi-review/INSTALL.md

@@ -5,10 +5,10 @@ This guide walks you through setting up the Delphi consensus review skill in you
 ## Prerequisites
 
 - OpenCode installed and configured
-- At least **2 different model providers** available (cross-provider requirement prevents homogenized blind spots)
-- Access to at least 3 different models (for 3-expert mode) or 2 models (for 2-expert mode)
+- Exactly **3 callable expert model configurations** are required.
+- The three configurations must use distinct trimmed executable model IDs. They may share one provider and token plan.
 
-> **Why cross-provider?** Using models from the same vendor (e.g., all OpenAI) means they share training data and biases, defeating the purpose of multi-expert consensus.
+> Provider, vendor, gateway, and model nationality are unrestricted. The enforced diversity boundary is three distinct requested model IDs, not provider diversity.
 
 ## Quick Setup (3 steps)
 
@@ -32,11 +32,8 @@ Then replace the provider/model placeholders:
 "model": "YOUR_PROVIDER/YOUR_MODEL_A"
 
 // After (your config):
-"model": "openai/gpt-4o"
-// or
-"model": "anthropic/claude-sonnet-4-20250514"
-// or
-"model": "bailian-coding-plan/qwen3.6-plus"
+"model": "bailian-tp/qwen-plus"
+// The other two role model fields must be different trimmed IDs.
 ```
 
 ### Step 3: Ensure provider configuration exists
@@ -73,7 +70,7 @@ For custom providers (like Ali Bailian), add a provider entry:
 
 ## Model Recommendations
 
-The skill requires at least 2 experts for code changes, 3 for architecture decisions. Here are recommended model selections:
+The skill requires exactly three experts for every mode. Choose three models that are callable through your configured provider.
 
 | Expert Role | Recommended | Alternatives |
 |-------------|-------------|-------------|
@@ -81,10 +78,11 @@ The skill requires at least 2 experts for code changes, 3 for architecture decis
 | **Technical (Expert B)** | Claude Haiku | Qwen-Coder, DeepSeek-Coder, GPT-4o-mini |
 | **Feasibility (Expert C)** | Claude Opus | GPT-4, Gemini 2.5 Pro, Qwen-Max |
 
-**Minimum viable setup** (2-expert mode):
-- Expert A: Any reasoning-strong model
-- Expert B: Any code-understanding-strong model
-- **Must be from different providers**
+**Minimum viable setup**:
+- Expert A: architecture model
+- Expert B: technical model
+- Expert C: feasibility model
+- All three requested model IDs must be distinct. A `provider: local` fallback does not count.
 
 ## Configuration File Reference
 
@@ -92,13 +90,13 @@ The skill requires at least 2 experts for code changes, 3 for architecture decis
 
 | Field | Description | Default |
 |-------|-------------|---------|
-| `num_experts` | Number of experts to use (2 or 3) | 3 |
+| `num_experts` | Number of experts to use | 3, fixed |
 | `experts.architecture` | Architecture reviewer configuration | Required |
 | `experts.technical` | Technical reviewer configuration | Required |
 | `experts.feasibility` | Feasibility reviewer configuration | Required for 3-expert mode |
     | `consensus.threshold_percent` | Agreement threshold | 90 |
 | `consensus.max_review_rounds` | Maximum review rounds | 5 |
-| `consensus.cross_provider_required` | Require different providers | true |
+| `consensus.cross_provider_required` | Deprecated and ignored; emits `cross_provider_required_ignored` | false or omitted |
 
 ### `opencode.json` agent block
 
@@ -133,11 +131,11 @@ You forgot to replace the placeholder. Search for `YOUR_PROVIDER` in your openco
 
 ### Both experts gave identical feedback
 
-Your models are from the same provider. Configure agents to use different providers.
+Shared providers are allowed. Configure three distinct executable model IDs and verify all three calls succeed.
 
 ### Review takes too long / costs too much
 
-Reduce to 2-expert mode in `.delphi-config.json`: set `num_experts: 2`.
+Do not reduce the expert count. Use three callable models and keep the threshold and round bounds unchanged.
 
 ## Advanced: JSON Schema Validation
 

package/skills/delphi-review/SKILL.md

@@ -4,8 +4,8 @@ version: 1.1.0
 description: >
   Performs multi-round anonymous expert consensus review using the Delphi method. Supports
   design review (default), code-walkthrough (--mode code-walkthrough), and requirements
-  (lightweight, --mode requirements) modes. Uses 2-3 experts from at least 2 different domestic
-  model providers with a >=90% statistical consensus threshold.
+  (lightweight, --mode requirements) modes. Uses exactly three experts with distinct executable
+  model IDs and a >=90% statistical consensus threshold.
   Outputs structured verdict (APPROVED/PASS_WITH_CAVEATS/REQUEST_CHANGES/BLOCKED).
 
   WHAT: Anonymous multi-expert review with iterative consensus building for designs, plans,
@@ -187,9 +187,9 @@ tools_denied:
 
 **In Scope:**
 - Multi-round anonymous expert consensus review (design + code-walkthrough modes)
-- 2-3 experts from different providers with statistical consensus (>= 90%)
+- Exactly three experts with distinct normalized executable model IDs and statistical consensus (>= 90%)
 - Structured verdict: APPROVED / PASS_WITH_CAVEATS / REQUEST_CHANGES
-- Domestic models only (no Anthropic/OpenAI/Google)
+- Model nationality, vendor, provider, and gateway are unrestricted.
 
 **Out of Scope:**
 - Does NOT implement code changes (review only, implementation is separate)
@@ -248,12 +248,12 @@ tools_denied:
 ## Workflow
 
 1. **Step 0: Input Validation** — Check input contains reviewable content (design doc/code/spec/diff). Empty input → `[DelphiReview:BLOCKED]`.
-2. **Round 1: Anonymous Independent Review** — Experts review without seeing each other's opinions. Output individual verdict JSON.
-3. **Consensus Check** — Consensus >=90% AND all APPROVED → complete.
-4. **Round 2: Exchange Opinions** — Experts see others' opinions, re-evaluate.
-5. **Round 3: Final Positions** — Final stance, output disagreements if any.
-6. **Fix & Re-Review** — REQUEST_CHANGES → fix Critical+Major → restart from Round 2.
-7. **Generate Output** — Consensus report + specification.yaml (design) or `.code-walkthrough-result.json` (walkthrough) + `delphi-reviewed.json`.
+2. **Round 1: Anonymous Independent Review** — Invoke architecture, technical, and feasibility independently without exposing their opinions to one another. Every successful result must use `result_type=delphi_expert_result`.
+3. **Execution Verification** — Verify that all three results succeeded and that their `requested_model` values are three distinct trimmed IDs. A local fallback (`provider: local`) is not an executed expert and cannot satisfy this check; a local hosted endpoint configured as an ordinary callable provider can.
+4. **Consensus Check** — Aggregate all three successful expert results. Consensus >=90% AND all APPROVED → complete. One expert result is never global approval.
+5. **Rounds 2-5: Exchange and Final Positions** — If needed, expose prior aggregate evidence, re-evaluate, and stop at the first approved consensus or after five rounds.
+6. **Failure Handling** — Any expert failure, missing result, duplicate model ID, or unverifiable execution blocks the review. Do not substitute a local fallback or silently reduce the expert count.
+7. **Generate Output** — Write the consensus report and mode-specific evidence only after aggregation. For code walkthrough, bind the exact-HEAD result after aggregation.
 
 ## Activation (MANDATORY for L1 Trigger Detection)
 
@@ -277,7 +277,7 @@ Each round MUST output a structured round marker:
 
 **Rules**:
 - Every round marker MUST appear as a separate, identifiable line
-- Round numbering MUST be sequential (1, 2, 3)
+- Round numbering MUST be sequential (1 through at most 5)
 - After each round, output consensus summary: `consensus_ratio=N/N`, `verdict_status: converging | stable | diverging`
 - Final round MUST output verdict: `APPROVED | PASS_WITH_CAVEATS | REQUEST_CHANGES | PROCESS_BLOCK`
 
@@ -316,16 +316,16 @@ Each round MUST output a structured round marker:
 
 | 配置 | 专家 | 适用场景 |
 |------|------|---------|
-| 2 专家（默认） | A(架构) + B(实现) | 代码变更、小型设计 |
-| 3 专家 | A(架构) + B(实现) + C(可行性) | 架构决策、需求文档 |
+| 3 专家（固定） | A(架构) + B(实现) + C(可行性) | 所有评审模式 |
 
 ### 模型选择策略（强制 — 平台适配）
 
 **关键原则**：
-- ✅ 三个专家必须来自 **至少 2 家不同 provider**
-- ❌ 禁止 hardcode 模型名称
-- ❌ 禁止三个专家全部使用同一 provider 的模型
-- ❌ 禁止使用 Anthropic/OpenAI/Google 等国外模型
+- ✅ 必须配置 architecture、technical、feasibility 三个专家角色
+- ✅ 三个角色必须使用三个不同的 trimmed、可执行模型 ID
+- ✅ 模型国籍、厂商、provider 和 gateway 不受限制
+- ✅ 三个角色可以共享一个 provider 和 token 计划
+- ❌ `provider: local` 的特殊 fallback 不能计为一次成功执行
 
 #### Qoder 平台（推荐 — Custom Agent 模式）
 
@@ -345,7 +345,7 @@ Qoder 环境下使用 **Custom Agent** 机制，每个专家是一个独立的 c
 
 OpenCode 环境下通过 `opencode.json` 的 agent 配置 + `.delphi-config.json` 调用外部 API：
 - **MUST 从 `opencode.json` 的 agent 配置中读取模型**
-- 通过 `scripts/delphi-external-review.cjs` 调用各 provider 的 OpenAI 兼容 API
+- 通过 `scripts/delphi-external-review.cjs` 调用各 provider 的兼容 API
 - 需要用户自行配置 API key（环境变量注入）
 
 ### 共识阈值
@@ -418,10 +418,7 @@ Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
 
 ## Output Format (MANDATORY for L3)
 
-**⚠️ Single vs Multi-Expert Output**:
-- **Multi-Expert Mode (default)**: MUST use the full JSON schema below. Each expert outputs independently; the orchestrator aggregates into `consensus_report`.
-- **Single Reviewer Mode** (explicit `--single`): MAY use simplified text template: `[DelphiReview] verdict=APPROVED|REQUEST_CHANGES|BLOCKED confidence=N/10 issues=[critical:N, major:N, minor:N] summary: [1-2 sentences]`.
-- **Never mix formats**.
+Every Delphi mode MUST use the full JSON schema below. Architecture, technical, and feasibility each output independently; the orchestrator verifies all three executions before aggregating `consensus_report`. There is no single-reviewer Delphi approval path.
 
 ```json
 {
@@ -451,13 +448,15 @@ Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
   "mode": "requirements",
   "verdict": "APPROVED|GAPS_FOUND",
   "timestamp": "2026-07-25T10:30:00Z",
+  "consensus_ratio": 1.0,
   "requirements_hash": "<SHA-256 hex digest>",
   "head_commit": "<git rev-parse HEAD>",
   "context_file_used": "CONTEXT.md",
   "round": 1,
   "expert_verdicts": [
-    { "role": "architecture", "verdict": "APPROVED", "confidence": 9 },
-    { "role": "feasibility", "verdict": "APPROVED", "confidence": 8 }
+    { "role": "architecture", "verdict": "APPROVED", "confidence": 9, "result_type": "delphi_expert_result", "requested_model": "provider/model-a" },
+    { "role": "technical", "verdict": "APPROVED", "confidence": 8, "result_type": "delphi_expert_result", "requested_model": "provider/model-b" },
+    { "role": "feasibility", "verdict": "APPROVED", "confidence": 8, "result_type": "delphi_expert_result", "requested_model": "provider/model-c" }
   ],
   "requirements_statement": "<short summary of what was reviewed>",
   "gaps_found": [],
@@ -466,6 +465,8 @@ Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
 }
 ```
 
+Schema-v2 Phase 2 requires `requirements_statement`, `timestamp`, `consensus_ratio`, `expert_verdicts`, `head_commit`, and `requirements_hash`. The hash is SHA-256 of the exact concatenation `requirements_statement + project-relative context file UTF-8 content if used + timestamp YYYY-MM-DD`; context paths must remain inside the project after realpath resolution.
+
 **Anti-patterns mapping:**
 - `Round 1 → "评审完成"` → MUST NOT have `verdict: APPROVED` if `critical_issues` exist
 - `只处理 Critical，忽略 Major` → MUST include `major_concerns` array
@@ -476,7 +477,7 @@ Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
 ## Terminal State Checklist
 
 - [ ] Phase 0 完成（文档验证 + 专家分配）
-- [ ] Round 1-3 完成（所有专家评审）
+- [ ] 已完成所需轮次（最多 5 轮，每轮三位专家均执行）
 - [ ] 问题共识比例 >=90%
 - [ ] 所有 Critical Issues 已解决，Major Concerns 已处理
 - [ ] 最终裁决是 **APPROVED** 或 **APPROVED_WITH_MINOR**
@@ -512,11 +513,11 @@ Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
 |---------|---------|
 | Round 1 未 APPROVED 就"评审完成" | 迭代直到 APPROVED，修复后重新评审 |
 | 只处理 Critical，忽略 Major | 零容忍：Critical/Major 全部必须处理 |
-| 单专家自评 | 至少 2 位不同 provider 的专家 |
+| 单专家自评 | 必须运行 architecture、technical、feasibility 三位专家，并聚合三份成功结果 |
 | 用户说"时间紧急"就跳过 | 评审是投资不是开销 |
 | "专家几乎一致"就通过 | "几乎" = 不一致，继续到 >=90% |
-| 使用 Anthropic/GPT/Gemini | 必须使用国产开源模型 |
-| 三个专家同一厂家 | 必须来自至少 2 家不同厂家 |
+| 使用任何国籍或厂商的模型 | 模型国籍和厂商均不受限制，仍必须满足三个 distinct executable model IDs |
+| 三个专家同一 provider | 允许共享 provider，只要三个 requested_model IDs distinct 且每个调用成功 |
 
 **Code-walkthrough 专属 Anti-Patterns** → 详见 `references/code-walkthrough.md`
 
@@ -531,7 +532,7 @@ Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
 | 要求跳过评审 | "skip review", "不用评审", "跳过评审" | → 提醒评审是投资而非开销 |
 | 时间压力 | "来不及", "时间紧", "emergency" | → 提醒时间紧迫正是需要评审的时刻 |
 | 提前终止 | Round 1 后用户说 "可以了", "够了" | → BLOCK: 评审未达终止条件 |
-| 单专家自评 | 用户仅指定 1 个专家 | → 提醒至少需要 2 位专家 |
+| 单专家自评 | 用户仅指定 1 个专家 | → BLOCK：Delphi 必须执行 architecture、technical、feasibility 三位专家 |
 | 无文档输入 | 仅触发词，无设计/代码内容 | → `[DelphiReview:BLOCKED]` |
 
 ---

package/skills/delphi-review/opencode.json.delphi.example

@@ -3,23 +3,19 @@
     "Delphi Review Agent Configuration Template",
     "===========================================",
     "Copy the relevant 'agent' block below into your opencode.json.",
-    "Replace YOUR_PROVIDER/YOUR_MODEL with your actual provider and model names.",
-    "At least 2 of the 3 agents should use different providers for cross-bias protection."
+    "Replace the example model IDs with three callable IDs available through your configured provider.",
+    "The three model fields must be different after trimming; providers may be shared."
   ],
 
   "$example_providers": {
-    "OpenAI": "openai/gpt-4o",
-    "Anthropic": "anthropic/claude-sonnet-4-20250514",
-    "Ali_Bailian": "bailian-coding-plan/qwen3.6-plus",
-    "Google_Gemini": "google/gemini-2.5-pro",
-    "DeepSeek": "deepseek/deepseek-chat"
+    "SharedProviderExample": "bailian-tp/qwen-plus"
   },
 
   "agent": {
     "delphi-reviewer-architecture": {
       "description": "Delphi expert reviewer for architecture perspective. Focuses on requirements alignment, system design, and high-level coherence.",
       "mode": "subagent",
-      "model": "YOUR_PROVIDER/YOUR_MODEL_A",
+      "model": "bailian-tp/qwen-plus",
       "prompt": "You are a Delphi expert reviewer specializing in architecture and requirements alignment. Review the input for: 1) Alignment with stated goals 2) Architectural coherence 3) System-level implications 4) Trade-offs and alternatives. Output structured review with verdict (APPROVED/REQUEST_CHANGES/REJECTED) and confidence level.",
       "tools": {
         "read": true,
@@ -31,7 +27,7 @@
     "delphi-reviewer-technical": {
       "description": "Delphi expert reviewer for technical depth. Focuses on implementation details, code correctness, and edge cases. Used in delphi-review.",
       "mode": "subagent",
-      "model": "YOUR_PROVIDER/YOUR_MODEL_B",
+      "model": "bailian-tp/deepseek-v3",
       "prompt": "You are a Delphi expert reviewer specializing in technical implementation. Review the input for: 1) Implementation correctness 2) Code quality and patterns 3) Edge cases and error handling 4) Performance implications. Output structured review with verdict (APPROVED/REQUEST_CHANGES/REJECTED) and confidence level.",
       "tools": {
         "read": true,
@@ -43,7 +39,7 @@
     "delphi-reviewer-feasibility": {
       "description": "Delphi expert reviewer for feasibility analysis. Focuses on practical constraints, risks, and execution viability. Used in delphi-review.",
       "mode": "subagent",
-      "model": "YOUR_PROVIDER/YOUR_MODEL_C",
+      "model": "bailian-tp/glm-4.5",
       "prompt": "You are a Delphi expert reviewer specializing in feasibility and risk analysis. Review the input for: 1) Practical constraints (time, resources, dependencies) 2) Risk identification and mitigation 3) Execution complexity 4) Alternative approaches. Output structured review with verdict (APPROVED/REQUEST_CHANGES/REJECTED) and confidence level.",
       "tools": {
         "read": true,

package/skills/delphi-review/references/code-walkthrough.md

@@ -14,11 +14,11 @@
 
 ## Five Core Properties
 
-1. **匿名性** — Expert A/B 互不知道对方意见
+1. **匿名性** — Expert A/B/C 在 Round 1 互不知道对方意见
 2. **迭代共识** — 多轮直到 APPROVED
 3. **零容忍** — Critical Issues 必须修复
 4. **零降级** — 环境/资源问题必须阻断，通知用户解决
-5. **强制覆盖** — 超过阈值必须 BLOCK，不能跳过
+5. **强制覆盖** — 变更必须完整评审；大型变更由用户选择完整评审或在评审前拆分
 
 ---
 
@@ -72,11 +72,13 @@ pre-push hook
     │         │
     │         ├─→ Expert B (实现 + 代码质量) 匿名评审 (含 principles_findings)
     │         │
+    │         ├─→ Expert C (可行性) 匿名评审 (含 principles_findings)
+    │         │
     │         ├─→ 共识检查
     │         │      │
-    │         │      ├─→ 2/2 APPROVED → 允许推送
+    │         │      ├─→ 三份成功结果、三个 distinct requested_model、≥90% → 允许推送
     │         │      │
-    │         │      ├─→ 分歧 → Expert C 仲裁
+    │         │      ├─→ 分歧 → 继续下一轮（最多 5 轮）
     │         │      │
     │         │      └─→ REQUEST_CHANGES → 阻塞推送
     │         │
@@ -93,9 +95,9 @@ pre-push hook
 |------|------|------|
 | Expert A | 架构 + 设计评审 | `.delphi-config.json` → `experts.architecture` |
 | Expert B | 实现 + 代码质量 | `.delphi-config.json` → `experts.technical` |
-| Expert C (仲裁) | 冲突裁决 | `.delphi-config.json` → `experts.feasibility` |
+| Expert C | 可行性评审 | `.delphi-config.json` → `experts.feasibility` |
 
-> ⚠️ **注意**: 至少配置 **两个不同 provider** 的模型。详见 [INSTALL.md](./INSTALL.md)。
+> 必须成功执行三个专家角色，并验证三个 distinct trimmed requested_model。Provider、vendor、gateway 和模型国籍均不受限制。`provider: local` fallback 不能计数，local hosted endpoint 可作为普通 callable provider。
 
 ---
 
@@ -171,8 +173,9 @@ pre-push hook
     │         │
     │         └─→ Step 2: Delphi Code Walkthrough
     │                  │
-    │                  ├─→ Expert A 评审 (含 principles_findings)
-    │                  ├─→ Expert B 评审 (含 principles_findings)
+    │                  ├─→ Expert A / architecture 评审 (含 principles_findings)
+    │                  ├─→ Expert B / technical 评审 (含 principles_findings)
+    │                  ├─→ Expert C / feasibility 评审 (含 principles_findings)
     │                  └─→ 共识检查
     │
     └─→ 允许/阻塞推送
@@ -237,10 +240,10 @@ pre-push hook
 
 | 条件 | 结果 |
 |------|------|
-| 2/2 APPROVED + 无 Critical Issues | ✅ 允许推送 |
-| 2/2 APPROVED + 有 Minor Issues | ✅ 允许推送（记录问题） |
-| 1/2 APPROVED | ⚠️ 需要 Expert C 仲裁 |
-| 0/2 APPROVED | ❌ 阻塞推送 |
+| 三份成功结果、三个 distinct requested_model、≥90% 且无 Critical Issues | ✅ 允许推送 |
+| 三份成功结果、≥90% 且有 Minor Issues | ✅ 允许推送（记录问题） |
+| 任一专家失败、缺失或模型重复 | ❌ 阻塞推送 |
+| 共识不足或有 Critical Issues | ❌ 阻塞推送 |
 | 有 Critical Issues | ❌ 阻塞推送 |
 | 有 Major Issues 且未处理 | ❌ 阻塞推送 |
 
@@ -251,26 +254,9 @@ pre-push hook
 | 指标 | 阈值 | 处理 |
 |------|------|------|
 | 单次走查成本 | ~$0.03 | 正常执行 |
-| 最大评审文件数 | 20 个文件 | 超过 → **BLOCK** |
-| 最大 diff 行数 | 500 行 | 超过 → **BLOCK** |
+| 变更规模 | 无硬性文件数或行数上限 | 大型变更必须完整评审或由用户选择拆分 |
 
-**超过阈值处理 (零降级)**：
-
-```
-IF 超过阈值:
-  → BLOCK 推送
-  → 通知用户：变更过大，无法有效评审
-  → 用户选项：
-     A. 拆分变更（推荐）
-     B. 用户明确授权跳过走查（需书面确认风险）
-
-  ❌ 禁止自动跳过走查
-  ❌ 禁止自动分批并批准
-```
-
-**设计原则**：
-- 大变更跳过评审 = 质量风险泄露到生产环境
-- 用户有权决定是否接受风险，AI 不能自动跳过
+大型变更没有自动跳过或绕过路径。用户可以选择完整评审，或在评审前主动拆分变更；每个变更仍须通过完整的走查和证据验证。
 
 ---
 
@@ -282,7 +268,7 @@ IF 超过阈值:
 BEFORE code walkthrough:
   ├─ 检查 Expert A 模型可用性
   ├─ 检查 Expert B 模型可用性
-  ├─ 检查 Expert C 模型可用性（仲裁时需要）
+  ├─ 检查 Expert C 模型可用性
   └─ 检查 OpenCode CLI 可用性
 
 IF 任何检查失败:
@@ -296,8 +282,7 @@ IF 任何检查失败:
 
 | 情况 | 处理 |
 |------|------|
-| Expert A API 错误 | BLOCK + 提示用户检查 API key |
-| Expert B API 错误 | BLOCK + 提示用户检查 API key |
+| 任一专家 API 错误 | BLOCK + 提示用户检查 provider/API key |
 | OpenCode CLI 缺失 | BLOCK + 提示用户安装 OpenCode |
 
 ---
@@ -305,7 +290,7 @@ IF 任何检查失败:
 ## Expert 评审输出格式
 
 ```markdown
-## Code Walkthrough - Expert [A/B]
+## Code Walkthrough - Expert [A/B/C]
 
 ### 变更摘要
 [变更的主要内容]
@@ -340,6 +325,7 @@ IF 任何检查失败:
 ### 专家意见汇总
 - Expert A: [APPROVED / REQUEST_CHANGES]
 - Expert B: [APPROVED / REQUEST_CHANGES]
+- Expert C: [APPROVED / REQUEST_CHANGES]
 
 ### 共识结果
 [CONSENSUS / DISAGREEMENT]
@@ -370,8 +356,9 @@ IF 任何检查失败:
   "verdict": "APPROVED",
   "confidence": 9,
   "experts": [
-    { "id": "Expert A", "verdict": "APPROVED", "confidence": 9 },
-    { "id": "Expert B", "verdict": "APPROVED", "confidence": 8 }
+    { "id": "Expert A", "role": "architecture", "verdict": "APPROVED", "confidence": 9, "result_type": "delphi_expert_result", "requested_model": "provider/model-a", "resolved_model": "provider/model-a" },
+    { "id": "Expert B", "role": "technical", "verdict": "APPROVED", "confidence": 8, "result_type": "delphi_expert_result", "requested_model": "provider/model-b", "resolved_model": null },
+    { "id": "Expert C", "role": "feasibility", "verdict": "APPROVED", "confidence": 8, "result_type": "delphi_expert_result", "requested_model": "provider/model-c", "resolved_model": "provider/model-c" }
   ],
   "issues": [],
   "consensus_ratio": 1.0,
@@ -409,7 +396,8 @@ Hook 验证以下条件（全部满足才允许 push）：
 2. JSON 格式有效
 3. commit hash 匹配当前 HEAD
 4. verdict = APPROVED
-5. timestamp 未过期 (< 1小时)
+5. timestamp 未过期
+6. expires 必须是 timestamp 之后恰好 1 小时
 
 ---
 
@@ -430,8 +418,7 @@ Hook 验证以下条件（全部满足才允许 push）：
 | 错误 | 正确 |
 |------|------|
 | Expert 不可用时跳过走查 | BLOCK + 提示用户修复环境 |
-| 超过阈值自动跳过走查 | BLOCK + 提示用户拆分变更 |
-| 超过阈值自动分批评审 | BLOCK + 用户决定是否拆分或授权跳过 |
+| 大型变更 | 完整评审或由用户在评审前拆分 |
 | OpenCode CLI 缺失时允许推送 | BLOCK + 提示用户安装 OpenCode |
 | 模型 API 错误时降级单模型 | BLOCK + 提示用户检查 API 配置 |
 | 未写入 `.code-walkthrough-result.json` 就声明完成 | MUST 写入结果文件后才能完成 |
@@ -443,15 +430,17 @@ Hook 验证以下条件（全部满足才允许 push）：
 
 **Pre-requisites (MANDATORY - BLOCK if missing):**
 - [ ] OpenCode CLI 已安装且可用
-- [ ] Expert A 模型 API 可用
-- [ ] Expert B 模型 API 可用
-- [ ] 变更大小在阈值内（文件 ≤20，行数 ≤500）
+- [ ] 三个专家模型 API 均可用
+- [ ] 变更已由用户选择完整评审或拆分为可审阅的变更
 - [ ] Expert A 完成匿名评审
 - [ ] Expert B 完成匿名评审
+- [ ] Expert C 完成匿名评审
 - [ ] 共识检查完成
 
 **CRITICAL - 共识验证 (code-walkthrough):**
-- [ ] 至少 2/2 或 2/3 专家 APPROVED
+- [ ] 三个成功结果均为 `result_type=delphi_expert_result`
+- [ ] 三个 `requested_model` trimmed 后 distinct
+- [ ] 聚合共识 >=90%，并在 exact HEAD 上写入结果
 - [ ] 无 Critical Issues 未解决
 - [ ] 无 Major Issues 未处理（或有处理计划）
 
@@ -471,11 +460,6 @@ Hook 验证以下条件（全部满足才允许 push）：
 - **CANNOT 允许推送**
 - **MUST 修复后重新走查**
 
-**IF 超过大小阈值:**
-- **CANNOT 允许推送**
-- **MUST BLOCK 并提示拆分变更**
-- **用户可明确授权跳过（需确认风险）**
-
 **⭐ APPROVED 后必做 (code-walkthrough mode)：写入结果文件**
 
 IF mode == `code-walkthrough` AND 最终裁决是 APPROVED，评审完成后必须：