xp-gate 0.5.1

package/skills/delphi-review/SKILL.md

@@ -0,0 +1,371 @@
+---
+name: delphi-review
+description: "Delphi consensus review: multi-round anonymous expert review until unanimous APPROVAL. Supports design/code-walkthrough modes. 2-3 experts from different providers. MANDATORY before implementation, design, or architecture decisions. Trigger: 'review this design', '评审这个需求', 'design review', '多专家评审', 'consensus review', 'code walkthrough', 'push review', or any request for multi-expert review of requirements, design docs, architecture, or PRs."
+---
+
+# Delphi Consensus Review
+
+## 核心原则
+
+**Delphi 方法只有一个目的：得到所有专家一致认可的可行方案。**
+
+### 四大核心特性（RAND 方法论）
+
+1. **匿名性** — Round 1 专家互不知道对方意见
+2. **迭代** — 多轮直到共识，不是固定轮数
+3. **受控反馈** — 每轮看到其他专家意见
+4. **统计共识** — >=95% 一致才算共识
+
+### 质量优先
+
+| 原则 | 说明 |
+|------|------|
+| Token 是投资 | 相比后期修复成本，评审消耗微不足道 |
+| APPROVED 才是终点 | REQUEST_CHANGES 必须修复并重新评审 |
+| 零容忍 | Critical/Major 问题全部必须处理，不可跳过或降级 |
+
+详见：Anti-Patterns 章节。
+
+---
+
+## 评审模式
+
+| 模式 | 触发 | 用途 | 输出 |
+|------|------|------|------|
+| `design`（默认） | `/delphi-review` | 需求/设计/架构/PR 评审 | 共识报告 + specification.yaml |
+| `code-walkthrough` | `/delphi-review --mode code-walkthrough` | git push 前代码走查 | `.code-walkthrough-result.json` |
+
+**Code Walkthrough 模式**的完整规范已移至 `references/code-walkthrough.md`。当用户使用 `--mode code-walkthrough` 时，读取该文件并执行其中定义的全部流程。
+
+---
+
+## 参数配置
+
+### 专家数量与角色
+
+| 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(可行性仲裁) | 架构决策、需求文档 |
+
+> ⚠️ 至少选择 **两家不同 provider** 的模型，避免同源盲点。模型映射见 `INSTALL.md`。
+> 
+> ### 模型选择策略（强制）
+> 
+> **MUST 使用国产开源模型**，严禁使用国外昂贵模型（Anthropic Claude、OpenAI GPT、Google Gemini 等）。
+> 
+> **允许的模型列表**（从以下国产开源模型厂家中选择）：
+> 
+> | 厂家 | 可选模型 |
+> |------|---------|
+> | 智谱 GLM | `glm-5.1`, `glm-5.0` |
+> | 月之暗面 Kimi | `kimi-k2.6`, `kimi-k2.5` |
+> | MiniMax | `minimax-m2.7`, `minimax-m2.5` |
+> | 阿里 Qwen | `qwen3.6-plus`, `qwen3.5-plus` |
+> | 深度求索 DeepSeek | `deepseek-v4-pro`, `deepseek-v4-lite` |
+> 
+> **专家分配建议**：
+> 
+> | Expert | 推荐模型 | 备选 |
+> |--------|---------|------|
+> | Expert A (架构) | `deepseek-v4-pro` | `qwen3.6-plus`, `glm-5.1` |
+> | Expert B (技术) | `kimi-k2.6` | `deepseek-v4-pro`, `minimax-m2.7` |
+> | Expert C (可行性) | `qwen3.6-plus` | `kimi-k2.6`, `glm-5.1` |
+> 
+> **关键原则**：
+> - ✅ 三个专家必须来自 **至少 2 家不同厂家**（如 DeepSeek + Kimi + Qwen）
+> - ❌ 禁止使用 Anthropic、OpenAI、Google 等国外模型
+> - ❌ 禁止三个专家全部使用同一厂家模型
+> - ⚠️ 成本控制：`deepseek-v4-lite`、`kimi-k2.5`、`qwen3.5-plus` 作为轻量备选
+
+### 共识阈值
+
+| 阈值 | 说明 |
+|------|------|
+| **>=95%** | 推荐默认 |
+| 100% | 完全一致（更严格） |
+
+---
+
+## 完整流程
+
+```
+Phase 0: 准备 → Round 1: 匿名独立评审 → 共识检查
+    │
+    ├─ 一致 + >=95% + APPROVED → ✅ 完成
+    │
+    └─ 不一致 或 <95% 或 REQUEST_CHANGES
+          │
+          ▼
+       Round 2: 交换意见 → 共识检查
+          │
+          ├─ 一致 + >=95% + APPROVED → ✅ 完成
+          │
+          └─ 仍分歧 或 REQUEST_CHANGES
+                │
+                ▼
+             Round 3: 最终立场 → 共识检查
+                │
+                ├─ APPROVED → ✅ 完成
+                │
+                └─ REQUEST_CHANGES → 修复方案 → 回到 Round 2 重新评审
+```
+
+---
+
+## Round 1: 匿名独立评审
+
+### 为什么必须匿名
+
+匿名防止 anchoring bias（锚定偏差）—— 知道其他专家意见后倾向于同意"权威"，不敢提出相反观点。
+
+### 执行方式
+
+每位专家独立收到：原始文档 + 评审模板 + "独立评审，不知道其他专家意见"。
+
+### 输出格式
+
+```markdown
+## 独立评审 - Expert [A/B/C]
+### 优点
+1. [具体优点 + 文档位置]
+### 问题清单
+#### Critical Issues (必须修复才能批准)
+1. [问题] - 位置: [...] - 修复建议: [...]
+#### Major Concerns (必须处理)
+1. [...]
+#### Minor Concerns (需要说明)
+1. [...]
+### 裁决: [APPROVED / REQUEST_CHANGES / REJECTED]
+### 置信度: [X/10]
+### 关键理由
+1. [...]
+```
+
+---
+
+## Round 2: 交换意见
+
+### 执行方式
+
+每位专家看到：原始文档 + 其他专家的评审 + "响应其他专家的关切，是否调整立场？"
+
+### 输出格式
+
+```markdown
+## Round 2 Response - Expert [A/B/C]
+### 响应其他专家关切
+**Expert [X] 提到: [问题]**
+- 我的立场: [同意/部分同意/不同意] - 理由: [...]
+### 更新后问题清单 / 裁决 / 置信度 / 立场变化说明
+```
+
+---
+
+## Round 3: 最终立场（如需要）
+
+触发条件：Round 2 后仍无共识。所有专家提交最终绑定立场。3 专家模式下若仍无完全一致，2/3 或 3/3 多数裁决生效，记录少数派意见。
+
+### 输出格式
+
+```markdown
+## Round 3 Final Position - Expert [A/B/C]
+### 最终裁决: [APPROVED / REQUEST_CHANGES / REJECTED]
+### 最终置信度: [X/10]
+### 关键理由 + 与其他专家的差异
+```
+
+---
+
+## 修复与重新评审
+
+如果最终裁决是 REQUEST_CHANGES 或 REJECTED：
+1. 修复所有 Critical Issues + 处理所有 Major Concerns
+2. 重新评审（从 Round 2 起步，不是 Round 1）
+3. 迭代直到 APPROVED
+
+修复报告格式：
+```markdown
+## 修复报告
+### Critical Issues 修复 | ### Major Concerns 处理 | ### Minor Concerns 说明
+### 请求重新评审
+```
+
+---
+
+## 终止条件
+
+| 参数 | 默认值 | 说明 |
+|------|--------|------|
+| `max_review_rounds` | 5 | 超过后生成"未达成共识报告"，交人决策 |
+| `timeout` | 60min | 单次评审超时 |
+
+---
+
+## Output Format (MANDATORY)
+Every review round output MUST follow this exact JSON structure:
+
+```json
+{
+  "expert_id": "A|B|C",
+  "round": 1,
+  "mode": "design|code-walkthrough",
+  "verdict": "APPROVED|REQUEST_CHANGES|REJECTED",
+  "confidence": 9,
+  "critical_issues": ["..."],
+  "major_concerns": ["..."],
+  "minor_concerns": ["..."],
+  "consensus_report": {
+    "agreed_items": ["..."],
+    "disagreed_items": ["..."],
+    "final_verdict": "APPROVED|REQUEST_CHANGES"
+  }
+}
+```
+
+**Anti-patterns mapping to assertions:**
+- `Round 1 → 生成报告 → "评审完成"` → Output MUST NOT have `verdict: APPROVED` if `critical_issues` exist.
+- `只处理 Critical，忽略 Major` → Output MUST include `major_concerns` array, even if empty.
+- `用户说"时间紧急"就跳过` → Output MUST include `round` field, proving multi-round process.
+
+---
+
+## Terminal State Checklist
+
+<MANDATORY-CHECKLIST>
+
+### 你只能在以下条件全部满足后声明"Delphi review complete":
+
+**Pre-requisites:**
+- [ ] Phase 0 完成（文档验证 + 专家分配）
+- [ ] Round 1 完成（所有专家匿名独立评审）
+- [ ] Round 2+ 完成（交换意见 / 最终立场）
+
+**CRITICAL — 共识验证:**
+- [ ] 问题共识比例 >=95%
+- [ ] 所有 Critical Issues 已解决
+- [ ] 所有 Major Concerns 已处理
+
+**CRITICAL — 裁决检查:**
+- [ ] 最终裁决是 **APPROVED** 或 **APPROVED_WITH_MINOR**
+- [ ] 如果 REQUEST_CHANGES → 已修复 → 已重新评审 → APPROVED
+
+**Final Requirements:**
+- [ ] 共识报告生成并保存
+- [ ] 用户已确认报告
+- [ ] ⭐ **IF APPROVED (design mode): 提示用户生成 specification.yaml，如用户同意则调用 specification-generator**
+
+**IF 裁决是 REQUEST_CHANGES 或 REJECTED → CANNOT claim complete, MUST 修复并重新评审**
+**IF 任何条件未满足 → CANNOT claim complete, MUST BLOCK 并通知用户**
+
+### ⭐ APPROVED 后必做
+
+**Design mode**:
+
+1. **Automatic: 生成 specification.yaml** — 从设计文档提取需求到 specification.yaml。这是 test-specification-alignment 的输入，没有它 test-spec 会进入不推荐的 legacy mode。
+
+2. **必须提示用户** — Delphi review APPROVED 后，agent 必须主动输出以下提示：
+
+   ```
+   ⭐ Delphi review APPROVED 完成！
+
+   Next Step: 生成 specification.yaml
+
+   设计文档已稳定，现在可以生成 specification.yaml 用于后续的 test-specification-alignment 验证。
+
+   是否现在生成？
+   - 回答 "yes" 或 "generate spec" → 我将调用 specification-generator 自动生成
+   - 回答 "no" 或 "later" → 稍后手动调用 /specification-generator
+   ```
+
+3. **如果用户同意** → 立即调用 `task(subagent_type="deep", load_skills=["specification-generator"], prompt="基于以下 APPROVED 设计文档生成 specification.yaml: [粘贴设计文档内容]")`
+
+4. **生成完成后** → 将 specification.yaml 包含在提交中，标记为 "auto-generated from Delphi review consensus"
+
+**Code-walkthrough mode**: 写入 `.code-walkthrough-result.json`（commit hash 匹配 HEAD，expires = timestamp + 1小时）。详见 `references/code-walkthrough.md`。
+
+### ⭐ 状态文件输出（MANDATORY — 防止跳过门禁）
+
+**Design mode APPROVED 后** — 必须立即写入 `.sprint-state/delphi-reviewed.json`：
+
+```json
+{
+  "mode": "design",
+  "timestamp": "2026-05-30T10:30:00Z",
+  "verdict": "APPROVED",
+  "consensus_ratio": 1.0,
+  "specification_path": ".sprint-state/phase-outputs/specification.yaml"
+}
+```
+
+**Code-walkthrough mode APPROVED 后** — 必须立即写入 `.sprint-state/delphi-reviewed.json`：
+
+```json
+{
+  "mode": "code-walkthrough",
+  "commit": "abc123def...",
+  "timestamp": "2026-05-30T10:30:00Z",
+  "verdict": "APPROVED",
+  "consensus_ratio": 1.0
+}
+```
+
+**用途**: Phase 2 BUILD 入口检查 (DELPHI-GATE) 读取此文件。`verdict != "APPROVED"` → 禁止编码。
+
+</MANDATORY-CHECKLIST>
+
+---
+
+## Anti-Patterns
+
+| ❌ 错误 | ✅ 正确 |
+|---------|---------|
+| Round 1 → 生成报告 → "评审完成"（未 APPROVED） | 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，不可跳过或降级 |
+| 单专家自评 | 至少 2 位不同 provider 的专家 |
+| 用户说"时间紧急"就跳过 | 评审是投资不是开销，跳过后期返工成本更高 |
+| "专家几乎一致"就通过 | "几乎" = 不一致，继续到 >=95% |
+| 使用 Anthropic/GPT/Gemini 等国外昂贵模型 | 必须使用国产开源模型（DeepSeek, Qwen, Kimi, GLM, MiniMax） |
+| 三个专家使用同一厂家模型 | 必须来自至少 2 家不同厂家 |
+
+**Code-walkthrough 专属 Anti-Patterns**: 详见 `references/code-walkthrough.md`。
+
+---
+
+## Red Flags
+
+| 借口 | 现实 |
+|------|------|
+| "这只是小变更" | 所有变更都需要评审 |
+| "Round 1 就够了" | 不够，必须多轮直到共识 |
+| "生成报告就完成了" | APPROVED 才算完成 |
+| "2/3 同意就是共识" | 还要检查问题共识比例 >=95% |
+
+---
+
+## 成功标准
+
+**Delphi 评审完成的唯一标准：**
+1. ✅ 所有专家裁决 APPROVED
+2. ✅ 问题共识 >=95%
+3. ✅ 所有 Critical Issues 已修复验证
+4. ✅ 所有 Major Concerns 已处理
+5. ✅ 共识报告已生成
+6. ✅ 用户已确认
+
+**缺少任何一项 = 未完成**
+## Output Format (MANDATORY)
+Every delphi review round MUST output valid JSON:
+```json
+{
+  "skill_name": "delphi-review",
+  "mode": "design|code-walkthrough",
+  "phase": "Round 1|Round 2|Round 3|Consensus",
+  "expert_id": "A|B|C",
+  "verdict": "APPROVED|REQUEST_CHANGES|REJECTED",
+  "confidence": 8,
+  "issues": [{"id": "string", "severity": "critical|major|minor", "description": "string"}],
+  "consensus_report": {"status": "pending|consensus|disagreement"}
+}
+```
+**Eval assertions check for:** `verdict` enum values, `confidence` range, `issues` structure, `consensus_report.status`.

package/skills/delphi-review/evals/evals.json

@@ -0,0 +1,82 @@
+{
+  "skill_name": "delphi-review",
+  "skill_path": "skills/delphi-review/SKILL.md",
+  "evals": [
+    {
+      "id": 1,
+      "name": "design-mode-full-review",
+      "category": "normal",
+      "prompt": "我需要评审一份设计文档：我们要给XP-Gate项目添加一个自动生成CHANGELOG的功能。功能包括：1) 从git log提取commit信息 2) 按照conventional commits分类 3) 生成markdown格式的changelog 4) 支持自定义模板。请用delphi-review评审这个设计。",
+      "expected_output": "执行完整的Delphi评审流程：Phase 0验证→Round 1匿名独立评审(至少2位Expert)→Round 2交换意见→共识检查(>=95%)→Terminal State Checklist→如果APPROVED则生成specification.yaml。绝不能只做Round 1就停止。",
+      "files": [],
+      "assertions": [
+        {"name": "round-1-executed", "type": "contains", "value": "Round 1"},
+        {"name": "round-2-executed", "type": "contains", "value": "Round 2"},
+        {"name": "consensus-check", "type": "contains", "value": "95%"},
+        {"name": "terminal-state-checklist", "type": "contains", "value": "Terminal State"},
+        {"name": "expert-a-present", "type": "contains", "value": "Expert A"},
+        {"name": "expert-b-present", "type": "contains", "value": "Expert B"},
+        {"name": "anonymity-mentioned", "type": "contains", "value": "匿名"},
+        {"name": "verdict-present", "type": "regex", "value": "(APPROVED|REQUEST_CHANGES|REJECTED)"}
+      ]
+    },
+    {
+      "id": 2,
+      "name": "design-mode-flawed-document",
+      "category": "boundary",
+      "prompt": "请评审以下架构决策：我们打算用单个SQLite文件存储所有用户数据、日志、审计记录和缓存。这样部署简单，只需要一个文件。请用delphi-review评审。",
+      "expected_output": "Expert应该识别出这个设计的严重问题（单文件SQLite不适合多用途、缺乏扩展性、并发问题等），给出REQUEST_CHANGES而非盲目APPROVED。仍需执行完整多轮流程。",
+      "files": [],
+      "assertions": [
+        {"name": "round-1-executed", "type": "contains", "value": "Round 1"},
+        {"name": "critical-issues-found", "type": "contains", "value": "Critical"},
+        {"name": "verdict-not-blind-approved", "type": "contains", "value": "REQUEST_CHANGES"}
+      ]
+    },
+    {
+      "id": 3,
+      "name": "code-walkthrough-mode",
+      "category": "normal",
+      "prompt": "我修改了几个文件准备push，需要做代码走查。变更包括：1) 新增了boy-scout.ts的calculateDelta函数 2) 修改了baseline.ts的loadBaseline函数 3) 添加了3个新的clean-code规则。请用code-walkthrough模式评审。",
+      "expected_output": "执行code-walkthrough模式：运行Principles Checker→Expert A架构评审(含principles_findings)→Expert B实现评审(含principles_findings)→共识检查→写入.code-walkthrough-result.json文件。",
+      "files": [],
+      "assertions": [
+        {"name": "code-walkthrough-mode", "type": "contains", "value": "code-walkthrough"},
+        {"name": "principles-findings", "type": "contains", "value": "Principles Findings"},
+        {"name": "expert-a-review", "type": "contains", "value": "Expert A"},
+        {"name": "expert-b-review", "type": "contains", "value": "Expert B"},
+        {"name": "result-file-generated", "type": "contains", "value": ".code-walkthrough-result.json"},
+        {"name": "consensus-check", "type": "contains", "value": "共识"}
+      ]
+    },
+    {
+      "id": 4,
+      "name": "zero-tolerance-enforcement",
+      "category": "boundary",
+      "prompt": "评审时发现了一个Critical Issue，但我觉得这个问题不大，可以忽略。评审还需要继续吗？",
+      "expected_output": "必须坚持零容忍原则：Critical Issues不可忽略，必须修复后重新评审。不能因为用户认为问题不大就跳过。应明确告知零容忍原则并拒绝跳过。",
+      "files": [],
+      "assertions": [
+        {"name": "zero-tolerance-enforced", "type": "contains", "value": "零容忍"},
+        {"name": "critical-not-skippable", "type": "not_contains", "value": "可以忽略"},
+        {"name": "requires-fix", "type": "contains", "value": "修复"}
+      ]
+    }
+  ],
+  "trigger_evals": {
+    "should_trigger": [
+      "我需要多个专家来评审这个架构设计",
+      "评审这份需求文档，确保没有遗漏",
+      "设计评审：我们的微服务拆分方案",
+      "代码走查，准备push了",
+      "delphi review this design doc"
+    ],
+    "should_not_trigger": [
+      "帮我写一个单元测试",
+      "这段代码有bug，帮我修一下",
+      "创建一个新的React组件",
+      "部署到生产环境",
+      "解释一下这个函数的作用"
+    ]
+  }
+}

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

@@ -0,0 +1,56 @@
+{
+  "$comment": [
+    "Delphi Review Agent Configuration Template",
+    "===========================================",
+    "Copy the relevant 'agent' block below into your opencode.json.",
+    "Replace YOUR_PROVIDER/YOUR_MODEL with your actual provider and model names.",
+    "At least 2 of the 3 agents should use different providers for cross-bias protection."
+  ],
+
+  "$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"
+  },
+
+  "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",
+      "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,
+        "bash": true,
+        "write": false,
+        "edit": false
+      }
+    },
+    "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",
+      "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,
+        "bash": true,
+        "write": false,
+        "edit": false
+      }
+    },
+    "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",
+      "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,
+        "bash": true,
+        "write": false,
+        "edit": false
+      }
+    }
+  }
+}