@pigcloud/skills 1.0.0

package/skills/domain-modeling/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: domain-modeling
+description: Extract domain boundaries and core concepts from code or business language when the model is unclear
+lifecycle_stage: design
+rule_profile: modeling
+dependencies: []
+triggers:
+  - domain
+  - boundary
+  - reverse engineering
+  - ubiquitous language
+  - module split
+  - model
+inputs:
+  - existing business language
+  - product context
+  - code or documents
+outputs:
+  - domain map
+  - core concepts
+  - boundary notes
+workflow:
+  - extract the core domain terms
+  - group concepts by boundary
+  - identify aggregates and dependencies
+  - hand off to technical-design or spec-refinement
+gates:
+  - stop at domain boundaries
+  - do not implement code
+  - do not write PRD text
+refs:
+  - skills/references/skill-reference-matrix.md
+  - skills/references/rule-loading-map.md
+  - skills/references/prompt-replay-checklist.md
+  - skills/references/full-chain-replay-scenarios.md
+---
+
+# Domain Modeling
+
+## Purpose
+
+Extract domain boundaries and core concepts from messy input.
+
+## Suitable / Unsuitable
+
+- Suitable: domain split, reverse engineering, ubiquitous language, boundary discovery
+- Unsuitable: implementation, PRD drafting, detailed test authoring
+
+## Workflow
+
+1. Extract the core domain terms.
+2. Group concepts by boundary.
+3. Identify aggregates and dependencies.
+4. Hand off to `technical-design` or `spec-refinement`.
+
+## Replay Signals
+
+- Input signal: existing business language, product context, code, or documents.
+- Output to verify: domain map, core concepts, boundary notes.
+- Stop signal: implementation code or PRD drafting.
+- Handoff signal: the boundary map is ready for `technical-design` or `spec-refinement`.
+
+## Examples
+
+- Input: "Reverse engineer the customer order domain from this codebase and docs."
+- Output: domain map, core concepts, boundary notes
+- Stop: do not write implementation code or PRD text
+
+## Stop Rules
+
+- Stop at domain boundaries.
+- Do not implement code.
+- Do not write PRD text.
+
+## References
+
+- `skills/references/skill-reference-matrix.md`
+- `skills/references/rule-loading-map.md`
+- `skills/references/prompt-replay-checklist.md`
+- `skills/references/full-chain-replay-scenarios.md`

package/skills/domain-modeling/references/README.md

@@ -0,0 +1,134 @@
+# 蒸馏产物使用说明
+
+> 如何将 `distilled/` 目录产物用于 AI 应用开发
+
+---
+
+## 作为 AI 应用开发者，如何把产物用起来
+
+### 1. 常驻上下文（语义层）
+
+把语义层注入系统提示或检索的高优先级片段：
+
+```
+- 00-overview.md — 系统全景与给 AI 的约定
+- 01-glossary.md — 业务术语
+- domain/_index.md — 对象关系总览
+- api/_index.md — API 能力清单与典型编排
+```
+
+**加载策略**：
+- 作为 System Prompt 的一部分
+- 或作为 RAG 检索的高优先级片段（权重 > 0.8）
+
+### 2. 按需检索（明细层）
+
+把明细层做成可被 AI 检索的知识库：
+
+```
+- api/<域>/<用例>.md — 各业务用例接口详情
+- reference/schema/*.md — 字段级 Schema
+- reference/enums.md — 枚举/字典值
+- reference/error-codes.md — 错误码
+- reference/model-code-conflicts.md — 模型与代码差异
+```
+
+**检索策略**：
+- 使用向量数据库（如 Milvus、Pinecone）
+- 按 `[[slug]]` 交叉引用定位
+- AI 需要时再取，不常驻上下文
+
+### 3. 接口对接
+
+按 `api/` 文档把真实 API 注册为 AI 可调用的工具：
+
+```json
+{
+  "tools": [
+    {
+      "name": "create_contract",
+      "description": "录入销售合同",
+      "parameters": {
+        "type": "object",
+        "properties": {
+          "contractNo": { "type": "string", "description": "合同编号" },
+          "contractName": { "type": "string", "description": "合同名称" },
+          ...
+        }
+      }
+    }
+  ]
+}
+```
+
+### 4. 运行时护栏
+
+把 `00-overview.md` 的"给 AI 的关键约定"写进系统提示：
+
+```
+## 给 AI 的关键约定
+1. 永远通过 API 操作，不直接读写数据库。
+2. 引用类字段是 ID 不是名称——先调主数据接口解析。
+3. 金额/状态等由服务端计算或裁决。
+4. 写操作权限分级：前期只读，写需确认。
+5. 规则标注 【服务端强制】/【调用方需预判】 的含义。
+```
+
+---
+
+## 写权限控制的落地建议
+
+### 前期阶段
+
+- 只把**查询类**接口注册为 AI 工具
+- 写接口不注册或需人工确认
+
+### 进阶阶段
+
+加一层**调用网关**：
+
+```
+GET/查询类 → 直接放行
+写操作 → dry-run 校验 → 回显 → 用户确认 → 提交
+```
+
+### 不可逆操作
+
+- 如确认收款无撤销
+- 必须强制二次确认
+
+---
+
+## 典型对话流程
+
+```
+用户: "给华东能源录个合同，软件平台产品，销售一部张敏负责，48 万"
+
+AI:
+1. 解析名称 → 调用 reference-data 接口获取 ID
+2. 回显解析结果给用户确认
+3. 用户确认后调用 create-contract 接口
+4. 返回创建结果
+```
+
+---
+
+## 错误处理
+
+当接口返回错误时，AI 根据 `error-codes.md` 解释：
+
+```
+错误: "付款比例合计必须等于 1"
+AI 解释: "您输入的付款比例合计不等于 100%，请调整各阶段比例"
+```
+
+---
+
+## 质量验证
+
+定期执行实跑验证：
+
+1. 准备测试集（读/写/统计/规则反例）
+2. 让 AI 仅凭蒸馏文档推导调用
+3. 真打到接口验证结果
+4. 修复发现的文档缺口

package/skills/domain-modeling/references/distillation-checklist.md

@@ -0,0 +1,152 @@
+# 蒸馏质量验收检查清单
+
+> 用于验证蒸馏产物是否符合规范
+
+---
+
+## 静态检查清单
+
+### 目录结构检查
+
+- [ ] `distilled/` 目录存在
+- [ ] `README.md` 存在且内容完整
+- [ ] `00-overview.md` 存在且包含"给 AI 的关键约定"
+- [ ] `01-glossary.md` 存在
+- [ ] `domain/_index.md` 存在
+- [ ] `domain/<对象>.md` 每个核心对象一个文件
+- [ ] `api/_index.md` 存在且包含典型编排
+- [ ] `api/<域>/<用例>.md` 每个用例一个文件
+- [ ] `reference/schema/*.md` 字段级 Schema
+- [ ] `reference/enums.md` 存在
+- [ ] `reference/error-codes.md` 存在
+- [ ] `reference/model-code-conflicts.md` 存在
+- [ ] `requirements/_index.md` 存在
+
+---
+
+### 领域对象检查
+
+- [ ] 每个对象有"一句话定义"
+- [ ] 每个对象有"业务语义"段落
+- [ ] 每个对象有"生命周期/状态"段落
+- [ ] 每个对象有"关键属性"表（语义级）
+- [ ] 每个对象有"组合/关系"说明
+- [ ] 每个对象有"核心业务规则"段落
+- [ ] 每条规则有执行方标注（【服务端强制】/【调用方需预判】等）
+- [ ] 每个对象有"相关能力"链接
+- [ ] 每个对象有"明细索引"
+
+---
+
+### API 接口检查
+
+- [ ] 每个接口有"业务场景"说明
+- [ ] 每个接口有"能力边界"说明（做什么/不做什么）
+- [ ] 每个接口有"输入"表（参数/类型/必填/说明/约束）
+- [ ] 每个接口有"输出"表（字段/说明）
+- [ ] 每个接口有"关键规则"段落
+- [ ] 每条规则有执行方标注
+- [ ] 每个接口有"涉及的数据对象"链接
+- [ ] 每个接口有"可能的错误"表
+- [ ] 每个接口有"调用示例"（真实可用）
+- [ ] 每个接口有"写权限分级"
+
+---
+
+### 明细层检查
+
+- [ ] Schema 表有"对应领域属性/API字段"列
+- [ ] 枚举表有 ✅/⚠️ 标记
+- [ ] 错误码表有"AI 应如何向用户解释"列
+- [ ] 冲突清单有"本体/需求"、"代码"、"AI 影响"三列
+
+---
+
+### 交叉引用检查
+
+- [ ] 所有链接使用 `[[slug]]` 格式
+- [ ] 对象↔接口↔Schema 三者互相链接
+- [ ] 无明显断链（链接目标不存在但未标记"待补")
+
+---
+
+### 命名规范检查
+
+- [ ] 文件名使用小写 kebab-case
+- [ ] API 字段沿用代码实际返回的命名
+- [ ] 枚举值使用业务实际值（含中文）
+
+---
+
+## 动态验证清单
+
+### 读/查询场景
+
+| 测试 | 验证内容 | 结果 |
+|------|----------|------|
+| 名称→ID 解析 | AI 能正确调用主数据接口解析名称 | ✅/❌ |
+| 查询参数 | AI 能正确构造查询参数 | ✅/❌ |
+| 结果解读 | AI 能正确解读返回结果 | ✅/❌ |
+
+### 写/录入场景
+
+| 测试 | 验证内容 | 结果 |
+|------|----------|------|
+| 必填字段 | AI 能识别必填字段并提示用户 | ✅/❌ |
+| 用例编排 | AI 能按正确顺序调用接口 | ✅/❌ |
+| 服务端维护字段 | AI 不传入服务端维护字段 | ✅/❌ |
+| 回显确认 | 写操作前回显给用户确认 | ✅/❌ |
+
+### 统计场景
+
+| 测试 | 验证内容 | 结果 |
+|------|----------|------|
+| 口径理解 | AI 能正确理解统计接口口径 | ✅/❌ |
+| 参数构造 | AI 能正确构造统计参数 | ✅/❌ |
+
+### 规则反例场景
+
+| 测试 | 验证内容 | 结果 |
+|------|----------|------|
+| 规则预判 | AI 能预判违反规则的错误 | ✅/❌ |
+| 错误解释 | AI 能根据 error-codes.md 解释错误 | ✅/❌ |
+
+---
+
+## 验证结论
+
+| 检查项 | 通过数 | 总数 | 通过率 |
+|--------|--------|------|--------|
+| 目录结构 | | 13 | |
+| 领域对象 | | 9 | |
+| API 接口 | | 10 | |
+| 明细层 | | 4 | |
+| 交叉引用 | | 3 | |
+| 命名规范 | | 3 | |
+| 动态验证 | | 12 | |
+
+**总通过率要求**: >= 90%
+
+---
+
+## 验证记录
+
+| 日期 | 验证人 | 通过率 | 备注 |
+|------|--------|--------|------|
+| | | | |
+
+---
+
+## 发现的问题
+
+1. 
+2. 
+3. 
+
+---
+
+## 修复计划
+
+| 问题 | 修复方案 | 状态 |
+|------|----------|------|
+| | | 待修复/已修复 |

package/skills/domain-modeling/references/test-cases-template.md

@@ -0,0 +1,128 @@
+# 实跑验证测试集模板
+
+> 用于 S6 验证阶段的测试集准备
+
+---
+
+## 测试集四类场景
+
+### 1. 读/查询场景
+
+| 测试ID | 用户输入 | 预期行为 | 验证点 |
+|--------|----------|----------|--------|
+| R01 | "查 X 部门今年的合同" | 调用 query-contracts 接口 | 名称→ID 解析、查询参数 |
+| R02 | "查看合同 C001 的详情" | 调用 get-contract-detail 接口 | ID 定位、详情字段解读 |
+| R03 | "列出所有待开票的合同" | 调用 query-open-invoices 接口 | 状态过滤、列表解读 |
+| R04 | "查客户 A 的所有合同" | 调用 query-contracts 接口 | 客户 ID 解析、过滤参数 |
+
+---
+
+### 2. 写/录入场景
+
+| 测试ID | 用户输入 | 预期行为 | 验证点 |
+|--------|----------|----------|--------|
+| W01 | "给客户 A 录个合同，48 万，税率 13%" | 调用 create-contract 接口 | ID 解析、必填字段、回显确认 |
+| W02 | "给合同 C001 开票，阶段 1，金额 14.4 万" | 调用 create-invoice 接口 | 合同详情获取、可开票额计算 |
+| W03 | "确认发票 I001 已收款" | 调用 receive-payment 接口 | 发票状态检查、收款确认 |
+| W04 | "作废合同 C001" | 调用 void-contract 接口（如有） | 状态检查、权限确认 |
+
+---
+
+### 3. 统计场景
+
+| 测试ID | 用户输入 | 预期行为 | 验证点 |
+|--------|----------|----------|--------|
+| S01 | "未来三个月各部门预计收多少" | 调用 receipt-budget 接口 | 统计口径理解、参数构造 |
+| S02 | "今年已开票总额是多少" | 调用统计接口 | 时间范围、聚合口径 |
+| S03 | "各部门合同数量分布" | 调用统计接口 | 分组维度、结果解读 |
+| S04 | "客户 A 的收款进度" | 调用 get-contract-detail 接口 | 累计收款计算、进度解读 |
+
+---
+
+### 4. 规则反例场景
+
+| 测试ID | 用户输入 | 预期错误 | 验证点 |
+|--------|----------|----------|--------|
+| E01 | "付款比例：预付 20% + 验收 30%" | "付款比例合计必须等于 1" | 规则预判、错误解释 |
+| E02 | "给已收款的发票再次确认收款" | "发票已收款，不可重复确认" | 状态检查、错误解释 |
+| E03 | "开票金额超过合同总额" | "开票金额超过剩余可开票额" | 额度计算、错误解释 |
+| E04 | "使用不存在的客户 ID" | "客户不存在" | ID 验证、错误解释 |
+
+---
+
+## 测试执行记录模板
+
+### 测试 R01
+
+| 字段 | 内容 |
+|------|------|
+| 测试ID | R01 |
+| 用户输入 | "查华东部门今年的合同" |
+| AI 推导 | 1. 调用 reference-data 解析"华东部门" → deptId=xxx<br>2. 调用 query-contracts(deptId=xxx, year=2026) |
+| 实际调用 | |
+| 返回结果 | |
+| 验证点 | 名称→ID 解析：✅/❌<br>查询参数：✅/❌ |
+| 结论 | 通过/失败 |
+| 备注 | |
+
+---
+
+### 测试 W01
+
+| 字段 | 内容 |
+|------|------|
+| 测试ID | W01 |
+| 用户输入 | "给星河制造录个合同，集成交付服务，销售二部李强负责，48 万，税率 13%，分预付 30%/验收 40%/尾款 30%" |
+| AI 推导 | 1. 调用 reference-data 解析客户/产品/部门/负责人<br>2. 回显解析结果<br>3. 用户确认<br>4. 调用 create-contract |
+| 实际调用 | |
+| 返回结果 | |
+| 验证点 | ID 解析：✅/❌<br>必填字段：✅/❌<br>回显确认：✅/❌<br>服务端维护字段未传入：✅/❌ |
+| 结论 | 通过/失败 |
+| 备注 | |
+
+---
+
+### 测试 E01
+
+| 字段 | 内容 |
+|------|------|
+| 测试ID | E01 |
+| 用户输入 | "付款比例：预付 20% + 验收 30%" |
+| AI 推导 | AI 应预判比例合计=0.5≠1，提示用户调整 |
+| 实际行为 | |
+| 返回结果 | |
+| 验证点 | 规则预判：✅/❌<br>错误解释：✅/❌ |
+| 结论 | 通过/失败 |
+| 备注 | |
+
+---
+
+## 验证汇总
+
+| 类型 | 测试数 | 通过数 | 通过率 |
+|------|--------|--------|--------|
+| 读/查询 | 4 | | |
+| 写/录入 | 4 | | |
+| 统计 | 4 | | |
+| 规则反例 | 4 | | |
+| **总计** | **16** | | |
+
+**判定标准**：
+- 通过率 >= 90% → 文档可用性通过
+- 通过率 < 90% → 需迭代修复
+
+---
+
+## 发现的文档缺口
+
+| 缺口 | 影响测试 | 修复方案 |
+|------|----------|----------|
+| | | |
+
+---
+
+## 迭代修复记录
+
+| 日期 | 修复内容 | 重新验证结果 |
+|------|----------|----------|
+| | | |

package/skills/environment-deploy/SKILL.md

@@ -0,0 +1,81 @@
+---
+name: environment-deploy
+description: Prepare deployment and runtime prerequisites when a workspace or target environment needs setup
+lifecycle_stage: scaffold
+rule_profile: scaffold
+dependencies:
+  - project-bootstrap
+triggers:
+  - deploy
+  - environment
+  - runtime
+  - install
+  - bootstrap
+  - setup
+inputs:
+  - deployment target
+  - runtime requirements
+  - workspace state
+outputs:
+  - environment plan
+  - deployment notes
+  - bootstrap instructions
+workflow:
+  - verify the target environment
+  - prepare runtime and deployment prerequisites
+  - document the setup steps
+  - hand off to feature-build or project-bootstrap
+gates:
+  - stop at environment setup
+  - do not change product scope
+  - do not generate business logic
+refs:
+  - skills/references/rule-loading-map.md
+  - skills/references/prompt-replay-checklist.md
+  - skills/references/full-chain-replay-scenarios.md
+  - rules/index.md
+---
+
+# Environment Deploy
+
+## Purpose
+
+Prepare deployment and runtime environments without changing application scope.
+
+## Suitable / Unsuitable
+
+- Suitable: runtime setup, deployment prerequisites, bootstrap instructions
+- Unsuitable: product discovery, architecture changes, implementation work
+
+## Workflow
+
+1. Verify the target environment.
+2. Prepare runtime and deployment prerequisites.
+3. Document the setup steps.
+4. Hand off to `feature-build` or `project-bootstrap`.
+
+## Replay Signals
+
+- Input signal: deployment target, runtime requirements, workspace state.
+- Output to verify: environment plan, deployment notes, bootstrap instructions.
+- Stop signal: product scope changes or business logic generation.
+- Handoff signal: the environment is ready for `feature-build` or `project-bootstrap`.
+
+## Examples
+
+- Input: "Prepare the runtime and deployment prerequisites for this workspace."
+- Output: environment plan, deployment notes, bootstrap instructions
+- Stop: do not change product scope or generate business logic
+
+## Stop Rules
+
+- Stop at environment setup.
+- Do not change product scope.
+- Do not generate business logic.
+
+## References
+
+- `skills/references/rule-loading-map.md`
+- `skills/references/prompt-replay-checklist.md`
+- `skills/references/full-chain-replay-scenarios.md`
+- `rules/index.md`

package/skills/feature-build/SKILL.md

@@ -0,0 +1,122 @@
+---
+name: feature-build
+description: Implement approved design changes when code, refactor, or integration work needs to stay within agreed boundaries
+lifecycle_stage: build
+rule_profile: implementation
+dependencies:
+  - technical-design
+triggers:
+  - coding
+  - implementation
+  - refactor
+  - entity
+  - controller
+  - service
+  - mapper
+inputs:
+  - executable spec
+  - design decisions
+  - confirmed task breakdown plan
+  - rules bundle
+outputs:
+  - code changes
+  - implementation notes
+  - validation evidence
+  - self-test evidence
+workflow:
+  - load the selected rules bundle
+  - confirm the task breakdown plan before coding
+  - implement one task at a time within the approved scope
+  - run local self-tests after each completed task
+  - record evidence for each finished task
+  - validate compile-time and local correctness
+  - hand off to test-design
+gates:
+  - stop at implementation
+  - do not rewrite the spec
+  - do not change the agreed boundaries
+  - do not bypass the rules bundle
+refs:
+  - skills/references/rule-loading-map.md
+  - skills/references/flow-test-cases.md
+  - skills/references/full-chain-replay-scenarios.md
+  - skills/references/prompt-replay-checklist.md
+  - rules/index.md
+---
+
+# Feature Build
+
+## Purpose
+
+Implement the approved design without changing the agreed scope.
+
+## Suitable / Unsuitable
+
+- Suitable: entity, mapper, service, controller, refactor, local validation, task-level implementation
+- Unsuitable: redesigning requirements, changing boundaries, bypassing rules, writing unrelated tests
+
+## Quick Start
+
+- Load the selected rules bundle first.
+- Implement only the confirmed scope from the approved design.
+- Split work into small tasks before implementation if the spec is still too coarse.
+- Treat the task as the smallest implementation unit.
+- Validate each task locally before starting the next one.
+- Record evidence immediately after each task is complete.
+- Validate compile-time and local correctness before handing off.
+
+## Inputs / Outputs
+
+- Inputs:
+  - executable spec
+  - design decisions
+  - confirmed task breakdown plan
+  - rules bundle
+- Outputs:
+  - code changes
+  - implementation notes
+  - validation evidence
+  - self-test evidence
+  - task-level evidence
+
+## Workflow
+
+1. Load the selected rules bundle.
+2. Confirm the task breakdown plan and smallest executable units.
+3. Implement one task at a time within the approved scope.
+4. Run local self-tests after each task and record the evidence.
+5. Validate compile-time and local correctness.
+6. Hand off to `test-design`.
+
+## Replay Signals
+
+- Input signal: confirmed spec, approved design, confirmed task breakdown plan, and selected rules bundle.
+- Output to verify: code changes, implementation notes, validation evidence, self-test evidence, task-level evidence.
+- Stop signal: spec rewrites, scope expansion, or bypassing the rules bundle.
+- Handoff signal: implementation is ready for `test-design`.
+
+## Gotchas
+
+- Do not rewrite the spec while implementing.
+- Do not expand the boundary because implementation is inconvenient.
+- Do not bypass the rules bundle for convenience.
+- Do not leave validation evidence implicit.
+- Do not batch unrelated tasks into a single implementation step.
+- Do not postpone self-test until the end of the whole change.
+- Do not lose task-level evidence.
+
+## Stop Rules
+
+- Stop at implementation.
+- Do not rewrite the spec.
+- Do not change the agreed boundaries.
+- Do not skip the rules bundle.
+- Do not bypass the rules bundle.
+
+## References
+
+- `skills/references/rule-loading-map.md`
+- `skills/references/flow-test-cases.md`
+- `skills/references/engineering-delivery-method.md`
+- `skills/references/engineering-delivery-template.md`
+- `rules/index.md`