@haaaiawd/anws 2.2.4 → 2.2.6
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +21 -26
- package/lib/manifest.js +2 -1
- package/package.json +1 -1
- package/templates/.agents/skills/anws-system/SKILL.md +108 -108
- package/templates/.agents/skills/code-reviewer/SKILL.md +8 -7
- package/templates/.agents/skills/e2e-testing-guide/SKILL.md +134 -135
- package/templates/.agents/skills/task-planner/SKILL.md +253 -699
- package/templates/.agents/skills/task-planner/references/TASK_TEMPLATE_05A.md +163 -0
- package/templates/.agents/skills/task-planner/references/TASK_TEMPLATE_05B.md +175 -0
- package/templates/.agents/skills/task-reviewer/SKILL.md +6 -6
- package/templates/.agents/workflows/blueprint.md +44 -358
- package/templates/.agents/workflows/challenge.md +40 -37
- package/templates/.agents/workflows/change.md +339 -346
- package/templates/.agents/workflows/craft.md +5 -12
- package/templates/.agents/workflows/design-system.md +674 -631
- package/templates/.agents/workflows/explore.md +399 -400
- package/templates/.agents/workflows/forge.md +90 -99
- package/templates/.agents/workflows/genesis.md +351 -353
- package/templates/.agents/workflows/probe.md +241 -243
- package/templates/.agents/workflows/quickstart.md +123 -123
- package/templates/.agents/workflows/upgrade.md +1 -11
- package/templates/AGENTS.md +149 -149
- package/templates/.agents/skills/task-planner/references/TASK_TEMPLATE.md +0 -214
package/README.md
CHANGED
|
@@ -4,7 +4,7 @@
|
|
|
4
4
|
<img src="assets/logo-cli.png" width="260" alt="Anws">
|
|
5
5
|
|
|
6
6
|
[](https://opensource.org/licenses/MIT)
|
|
7
|
-
[](https://github.com/Haaaiawd/ANWS/releases)
|
|
8
8
|
[](https://github.com/Haaaiawd/ANWS)
|
|
9
9
|
|
|
10
10
|
[English](./README.md) | [中文](./README_CN.md)
|
|
@@ -36,33 +36,28 @@ Anws enforces design-first principles, preserves context in files, and prevents
|
|
|
36
36
|
|
|
37
37
|
## Why Anws Exists
|
|
38
38
|
|
|
39
|
-
|
|
39
|
+
Many so-called spec systems are good at one thing: generating documents up front. That is not the hard part. The hard part is keeping implementation, verification, review, and upgrades aligned with those documents after the initial planning pass.
|
|
40
40
|
|
|
41
|
-
|
|
42
|
-
- different sessions generate incompatible structures
|
|
43
|
-
- **Context amnesia**
|
|
44
|
-
- a fresh chat loses system decisions, trade-offs, and task state
|
|
45
|
-
- **Planning collapse**
|
|
46
|
-
- code gets written before requirements and interfaces are stabilized
|
|
47
|
-
- **Unsafe upgrades**
|
|
48
|
-
- workflow files change over time, but existing projects cannot be updated cleanly
|
|
41
|
+
That is where AI coding usually breaks down:
|
|
49
42
|
|
|
50
|
-
|
|
43
|
+
- docs and code drift apart until nobody is sure which one is the real source of truth
|
|
44
|
+
- a new session loses context and starts guessing about boundaries, trade-offs, and task state
|
|
45
|
+
- code starts before requirements and interfaces are stable, so rework becomes inevitable
|
|
46
|
+
- automation increases speed, but review discipline and evidence quality quietly drop
|
|
47
|
+
- template upgrades become risky because existing projects cannot absorb new rules cleanly
|
|
51
48
|
|
|
52
|
-
|
|
53
|
-
- **A root anchor file** via `AGENTS.md`
|
|
54
|
-
- **Workflow-first execution** instead of prompt-only improvisation
|
|
55
|
-
- **Controlled update semantics** for `AGENTS.md`, installed targets, and upgrade history
|
|
49
|
+
Anws is not just a system for writing specs. It is a framework for keeping **design, execution, verification, review, and upgrade** tied together.
|
|
56
50
|
|
|
57
|
-
|
|
58
|
-
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
|
|
62
|
-
|
|
63
|
-
|
|
64
|
-
|
|
65
|
-
|
|
51
|
+
- **Documents are operating state, not attachments**
|
|
52
|
+
- PRD, Architecture, ADR, System Design, `05A_TASKS.md`, `05B_VERIFICATION_PLAN.md`, and `AGENTS.md` are part of the workflow runtime, not disposable planning notes.
|
|
53
|
+
- **Highly documented, but for control rather than ceremony**
|
|
54
|
+
- Design, task, and verification are separated on purpose so task ownership, validation responsibility, and evidence output stay traceable.
|
|
55
|
+
- **Automation stays fast without becoming sloppy**
|
|
56
|
+
- `/forge AUTO` speeds delivery up, but it does not bypass challenge pressure, code review, validation gates, or evidence closure.
|
|
57
|
+
- **It orchestrates verification, not only implementation**
|
|
58
|
+
- `05A` owns execution flow and `05B` owns verification planning. E2E, for example, is declared as a trigger assumption during planning and executed later in `/forge`, instead of being hand-waved as “already tested.”
|
|
59
|
+
- **Upgrades are controlled, not overwrite-driven**
|
|
60
|
+
- Templates, projections, and `AGENTS.md` have explicit update semantics, so existing projects can evolve without being effectively reinstalled from scratch.
|
|
66
61
|
|
|
67
62
|
---
|
|
68
63
|
|
|
@@ -127,7 +122,7 @@ What using Anws looks like in practice: architecture-first **`/genesis`**, **hum
|
|
|
127
122
|
## Philosophy
|
|
128
123
|
|
|
129
124
|
**1. Docs first—specs keep you in command**
|
|
130
|
-
PRD, architecture, tasks, and design land in the repo before code does—so the project doesn’t drift in aimless “vibe runs.” Scope and progress live in `.anws/`, `
|
|
125
|
+
PRD, architecture, tasks, and design land in the repo before code does—so the project doesn’t drift in aimless “vibe runs.” Scope and progress live in `.anws/`, `05A_TASKS.md`, `05B_VERIFICATION_PLAN.md`, and **`AGENTS.md`**: you stay **in control of the system**, not whichever chat window is open.
|
|
131
126
|
|
|
132
127
|
**2. Full autonomy inside the rails**
|
|
133
128
|
**`/forge` AUTO** is delegation with checkpoints: keep moving inside agreed contracts. **Code review**, **e2e-testing-guide**, and the rest of the template gates keep runs **auditable** and **bounded**. When a wave is executing, it’s reasonable to **walk away—coffee, a walk**—because confidence comes from the spec and gates, not from staring at the model.
|
|
@@ -149,7 +144,7 @@ Use Anws as a lifecycle, not just a folder pack.
|
|
|
149
144
|
| `/probe` | Analyze a legacy codebase before change | Existing code | Risk report |
|
|
150
145
|
| `/design-system` | Design one system in depth | Architecture overview | System design doc |
|
|
151
146
|
| `/challenge` | Review design, tasks, and implementation fidelity with adversarial pressure | Docs / tasks / code | Challenge report |
|
|
152
|
-
| `/blueprint` | Break architecture into executable work | PRD + architecture | `
|
|
147
|
+
| `/blueprint` | Break architecture into executable work | PRD + architecture | `05A_TASKS.md` + `05B_VERIFICATION_PLAN.md` |
|
|
153
148
|
| `/forge` | Turn approved tasks into code with challenge-report and contract gates | Tasks + review state | Working implementation |
|
|
154
149
|
| `/change` | In-version task/contract tweaks (controlled expansion: few new tasks) | Small scoped change | Updated task/design docs |
|
|
155
150
|
| `/explore` | Research ambiguous or strategic topics | Topic | Exploration report |
|
package/lib/manifest.js
CHANGED
|
@@ -44,7 +44,8 @@ const RESOURCE_REGISTRY = [
|
|
|
44
44
|
{ id: 'system-designer-detail-template', type: 'skill', source: '.agents/skills/system-designer/references/system-design-detail-template.md', fileName: 'system-designer/references/system-design-detail-template.md' },
|
|
45
45
|
{ id: 'system-designer-template', type: 'skill', source: '.agents/skills/system-designer/references/system-design-template.md', fileName: 'system-designer/references/system-design-template.md' },
|
|
46
46
|
{ id: 'task-planner', type: 'skill', source: '.agents/skills/task-planner/SKILL.md', fileName: 'task-planner/SKILL.md' },
|
|
47
|
-
{ id: 'task-planner-template', type: 'skill', source: '.agents/skills/task-planner/references/
|
|
47
|
+
{ id: 'task-planner-template-05a', type: 'skill', source: '.agents/skills/task-planner/references/TASK_TEMPLATE_05A.md', fileName: 'task-planner/references/TASK_TEMPLATE_05A.md' },
|
|
48
|
+
{ id: 'task-planner-template-05b', type: 'skill', source: '.agents/skills/task-planner/references/TASK_TEMPLATE_05B.md', fileName: 'task-planner/references/TASK_TEMPLATE_05B.md' },
|
|
48
49
|
{ id: 'task-reviewer', type: 'skill', source: '.agents/skills/task-reviewer/SKILL.md', fileName: 'task-reviewer/SKILL.md' },
|
|
49
50
|
{ id: 'tech-evaluator', type: 'skill', source: '.agents/skills/tech-evaluator/SKILL.md', fileName: 'tech-evaluator/SKILL.md' },
|
|
50
51
|
{ id: 'tech-evaluator-adr-template', type: 'skill', source: '.agents/skills/tech-evaluator/references/ADR_TEMPLATE.md', fileName: 'tech-evaluator/references/ADR_TEMPLATE.md' },
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@haaaiawd/anws",
|
|
3
|
-
"version": "2.2.
|
|
3
|
+
"version": "2.2.6",
|
|
4
4
|
"description": "Anws — A spec-driven workflow framework for AI-assisted development. Empowers prompt engineers to build production-ready software through structured PRD → Architecture → Task decomposition. Works with Claude Code, GitHub Copilot, Cursor, Windsurf, and any tool that reads AGENTS.md.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"anws",
|
|
@@ -1,108 +1,108 @@
|
|
|
1
|
-
---
|
|
2
|
-
name: anws-system
|
|
3
|
-
description: 当用户在 skills-only 环境中需要判断应该从哪个 anws 工作流开始,或需要在 forge / change / genesis / probe / blueprint / challenge / upgrade 之间路由时使用。它是 anws 工作流集合的导航入口。
|
|
4
|
-
---
|
|
5
|
-
|
|
6
|
-
# ANWS System Router Manual
|
|
7
|
-
|
|
8
|
-
你是 **ANWS Router**。
|
|
9
|
-
|
|
10
|
-
你的职责不是直接替代所有工作流,而是作为 **skills-only 模式** 下的统一导航入口:
|
|
11
|
-
|
|
12
|
-
- 判断当前请求应路由到哪个工作流
|
|
13
|
-
- 告诉模型还需要读取哪个 `references/*.md`
|
|
14
|
-
- 在开始执行前明确权限边界,避免把 `/forge`、`/change`、`/genesis` 混用
|
|
15
|
-
|
|
16
|
-
## 激活规则
|
|
17
|
-
|
|
18
|
-
出现以下任一情况时使用本 Skill:
|
|
19
|
-
|
|
20
|
-
1. 用户不知道该从哪个工作流开始
|
|
21
|
-
2. 用户明确提到 `/quickstart`、`/forge`、`/change`、`/genesis` 等工作流,但当前环境只有 skills
|
|
22
|
-
3. 用户请求涉及“下一步该走哪个流程”
|
|
23
|
-
4. 用户请求跨越设计、任务、实现多个阶段,需要先判断阶段边界
|
|
24
|
-
|
|
25
|
-
## 首次激活的强制步骤
|
|
26
|
-
|
|
27
|
-
1. 先读取 `references/quickstart.md`
|
|
28
|
-
2. 判断当前请求更接近哪一种场景
|
|
29
|
-
3. 再按需读取对应 workflow reference
|
|
30
|
-
4. 没有读完对应 reference 前,不得直接执行该 workflow 的写操作
|
|
31
|
-
|
|
32
|
-
## Workflow Map
|
|
33
|
-
|
|
34
|
-
- `references/quickstart.md`
|
|
35
|
-
- 用途:总入口。用于判断项目目前处于哪一阶段,以及应先调用哪个工作流
|
|
36
|
-
- `references/probe.md`
|
|
37
|
-
- 用途:接手遗留项目、重大改动前做系统风险探测
|
|
38
|
-
- `references/genesis.md`
|
|
39
|
-
- 用途:新项目、重大重构、架构升级、需要新版本时使用
|
|
40
|
-
- `references/design-system.md`
|
|
41
|
-
- 用途:为单个系统补详细设计文档
|
|
42
|
-
- `references/blueprint.md`
|
|
43
|
-
- 用途:将架构设计拆成可执行的 `
|
|
44
|
-
- `references/challenge.md`
|
|
45
|
-
- 用途:在编码前(或需要时含 `src/`)对抗式审查;可组合 design-reviewer、task-reviewer、**code-reviewer**(`CODE` / `FULL` 或自适应升级)
|
|
46
|
-
- `references/forge.md`
|
|
47
|
-
- 用途:按 `
|
|
48
|
-
- `references/change.md`
|
|
49
|
-
- 用途:在当前版本前提不变时微调任务/契约/验证承接;允许 **受控扩展** 下与用户原话或 `/forge` 回流对应的 **少量新任务**;**禁止**回填 `- [x]`、**禁止**运行或替代 **`code-reviewer`**
|
|
50
|
-
- `references/explore.md`
|
|
51
|
-
- 用途:做调研、探索、方案发散与收敛
|
|
52
|
-
- `references/craft.md`
|
|
53
|
-
- 用途:创建 workflow / skill / prompt;长模板、防护写法与自检清单在 **`craft-authoring`** skill(与 `/craft` 配套,路径随 target 投影到 `skills/craft-authoring/SKILL.md`)
|
|
54
|
-
- `references/upgrade.md`
|
|
55
|
-
- 用途:处理 `anws update` 之后的升级编排
|
|
56
|
-
|
|
57
|
-
## 路由规则
|
|
58
|
-
|
|
59
|
-
### Route 1: 不确定起点
|
|
60
|
-
|
|
61
|
-
如果用户不知道从哪里开始,或你对当前阶段没有把握:
|
|
62
|
-
|
|
63
|
-
1. 读取 `references/quickstart.md`
|
|
64
|
-
2. 根据项目状态判断入口
|
|
65
|
-
3. 给出建议 workflow,并说明理由
|
|
66
|
-
4. 等待用户确认
|
|
67
|
-
|
|
68
|
-
### Route 2: 请求是“开始编码 / 继续实现 / 做当前波次”
|
|
69
|
-
|
|
70
|
-
1. 读取 `references/forge.md`
|
|
71
|
-
2. 检查 `.anws/v{N}/
|
|
72
|
-
3. 若缺任务清单,不得直接实现,先回到 `blueprint` 或 `genesis`
|
|
73
|
-
4. 若任务含 **E2E / 浏览器手动验证**:在执行路径上读取 **`e2e-testing-guide`** skill(同目录 `skills/e2e-testing-guide/SKILL.md`);投影环境下路径以目标 IDE 的 `skills/` 为准
|
|
74
|
-
5. 在提交前需要静态契约核对时:读取 **`code-reviewer`** skill。**若宿主支持子代理**(如 Task、并行子会话等)→ **优先委派子代理**按该 skill 专职执行(隔离上下文,输出结构以 skill 为准)。**若无子代理能力** → 由**当前会话**按同一 skill 完整执行(检查清单、证据与输出要求不得缩水)。
|
|
75
|
-
|
|
76
|
-
### Route 3: 请求是“微调现有任务 / 修正文案 / 调整验收标准 / 受控补任务”
|
|
77
|
-
|
|
78
|
-
1. 读取 `references/change.md`
|
|
79
|
-
2. 判断变更是否仍属 `/change` 权限(含 **Q8 少量新任务**、契约/验证补全);若已改动需求/架构/ADR **前提** → `genesis`
|
|
80
|
-
3. **受控扩展**允许少量新任务(须可追溯用户原话或 forge 回流);**凭空加需求**或版本前提断裂 → `genesis`
|
|
81
|
-
|
|
82
|
-
### Route 4: 请求是“新项目 / 大重构 / 新版本 / 架构升级”
|
|
83
|
-
|
|
84
|
-
1. 读取 `references/genesis.md`
|
|
85
|
-
2. 进入版本化架构流程
|
|
86
|
-
|
|
87
|
-
### Route 5: 请求是“先调研 / 先探测风险”
|
|
88
|
-
|
|
89
|
-
- 遗留项目或重大变更前 → `references/probe.md`
|
|
90
|
-
- 技术调研或方案发散 → `references/explore.md`
|
|
91
|
-
|
|
92
|
-
## 边界守则
|
|
93
|
-
|
|
94
|
-
1. 你是导航层,不是所有 workflow 的替身
|
|
95
|
-
2. 每次只路由到一个主 workflow;如需串联,必须说明顺序
|
|
96
|
-
3. 未读取目标 workflow reference 前,不得假装已经遵循该流程
|
|
97
|
-
4. 若用户请求同时跨越设计与实现,先收敛边界,再决定 `/change` 或 `/genesis`
|
|
98
|
-
5. skills-only 模式下,workflow 详情全部位于 `references/`
|
|
99
|
-
|
|
100
|
-
## 输出格式
|
|
101
|
-
|
|
102
|
-
当你完成路由判断时,输出应包含:
|
|
103
|
-
|
|
104
|
-
- 当前判断的阶段
|
|
105
|
-
- 建议读取的 reference
|
|
106
|
-
- 建议进入的 workflow
|
|
107
|
-
- 为什么不是其他 workflow
|
|
108
|
-
- 是否需要等待用户确认
|
|
1
|
+
---
|
|
2
|
+
name: anws-system
|
|
3
|
+
description: 当用户在 skills-only 环境中需要判断应该从哪个 anws 工作流开始,或需要在 forge / change / genesis / probe / blueprint / challenge / upgrade 之间路由时使用。它是 anws 工作流集合的导航入口。
|
|
4
|
+
---
|
|
5
|
+
|
|
6
|
+
# ANWS System Router Manual
|
|
7
|
+
|
|
8
|
+
你是 **ANWS Router**。
|
|
9
|
+
|
|
10
|
+
你的职责不是直接替代所有工作流,而是作为 **skills-only 模式** 下的统一导航入口:
|
|
11
|
+
|
|
12
|
+
- 判断当前请求应路由到哪个工作流
|
|
13
|
+
- 告诉模型还需要读取哪个 `references/*.md`
|
|
14
|
+
- 在开始执行前明确权限边界,避免把 `/forge`、`/change`、`/genesis` 混用
|
|
15
|
+
|
|
16
|
+
## 激活规则
|
|
17
|
+
|
|
18
|
+
出现以下任一情况时使用本 Skill:
|
|
19
|
+
|
|
20
|
+
1. 用户不知道该从哪个工作流开始
|
|
21
|
+
2. 用户明确提到 `/quickstart`、`/forge`、`/change`、`/genesis` 等工作流,但当前环境只有 skills
|
|
22
|
+
3. 用户请求涉及“下一步该走哪个流程”
|
|
23
|
+
4. 用户请求跨越设计、任务、实现多个阶段,需要先判断阶段边界
|
|
24
|
+
|
|
25
|
+
## 首次激活的强制步骤
|
|
26
|
+
|
|
27
|
+
1. 先读取 `references/quickstart.md`
|
|
28
|
+
2. 判断当前请求更接近哪一种场景
|
|
29
|
+
3. 再按需读取对应 workflow reference
|
|
30
|
+
4. 没有读完对应 reference 前,不得直接执行该 workflow 的写操作
|
|
31
|
+
|
|
32
|
+
## Workflow Map
|
|
33
|
+
|
|
34
|
+
- `references/quickstart.md`
|
|
35
|
+
- 用途:总入口。用于判断项目目前处于哪一阶段,以及应先调用哪个工作流
|
|
36
|
+
- `references/probe.md`
|
|
37
|
+
- 用途:接手遗留项目、重大改动前做系统风险探测
|
|
38
|
+
- `references/genesis.md`
|
|
39
|
+
- 用途:新项目、重大重构、架构升级、需要新版本时使用
|
|
40
|
+
- `references/design-system.md`
|
|
41
|
+
- 用途:为单个系统补详细设计文档
|
|
42
|
+
- `references/blueprint.md`
|
|
43
|
+
- 用途:将架构设计拆成可执行的 `05A_TASKS.md`(任务)与 `05B_VERIFICATION_PLAN.md`(验证)
|
|
44
|
+
- `references/challenge.md`
|
|
45
|
+
- 用途:在编码前(或需要时含 `src/`)对抗式审查;可组合 design-reviewer、task-reviewer、**code-reviewer**(`CODE` / `FULL` 或自适应升级)
|
|
46
|
+
- `references/forge.md`
|
|
47
|
+
- 用途:按 `05A_TASKS.md` 执行编码,并读取 `05B_VERIFICATION_PLAN.md` 落实验证;在验证与提交之间调用 **code-reviewer**(静态忠实度)及按需 **`e2e-testing-guide`**(浏览器/E2E 指南或实测)
|
|
48
|
+
- `references/change.md`
|
|
49
|
+
- 用途:在当前版本前提不变时微调任务/契约/验证承接;允许 **受控扩展** 下与用户原话或 `/forge` 回流对应的 **少量新任务**;**禁止**回填 `- [x]`、**禁止**运行或替代 **`code-reviewer`**
|
|
50
|
+
- `references/explore.md`
|
|
51
|
+
- 用途:做调研、探索、方案发散与收敛
|
|
52
|
+
- `references/craft.md`
|
|
53
|
+
- 用途:创建 workflow / skill / prompt;长模板、防护写法与自检清单在 **`craft-authoring`** skill(与 `/craft` 配套,路径随 target 投影到 `skills/craft-authoring/SKILL.md`)
|
|
54
|
+
- `references/upgrade.md`
|
|
55
|
+
- 用途:处理 `anws update` 之后的升级编排
|
|
56
|
+
|
|
57
|
+
## 路由规则
|
|
58
|
+
|
|
59
|
+
### Route 1: 不确定起点
|
|
60
|
+
|
|
61
|
+
如果用户不知道从哪里开始,或你对当前阶段没有把握:
|
|
62
|
+
|
|
63
|
+
1. 读取 `references/quickstart.md`
|
|
64
|
+
2. 根据项目状态判断入口
|
|
65
|
+
3. 给出建议 workflow,并说明理由
|
|
66
|
+
4. 等待用户确认
|
|
67
|
+
|
|
68
|
+
### Route 2: 请求是“开始编码 / 继续实现 / 做当前波次”
|
|
69
|
+
|
|
70
|
+
1. 读取 `references/forge.md`
|
|
71
|
+
2. 检查 `.anws/v{N}/05A_TASKS.md` 与 `.anws/v{N}/05B_VERIFICATION_PLAN.md` 是否存在且已定义
|
|
72
|
+
3. 若缺任务清单,不得直接实现,先回到 `blueprint` 或 `genesis`
|
|
73
|
+
4. 若任务含 **E2E / 浏览器手动验证**:在执行路径上读取 **`e2e-testing-guide`** skill(同目录 `skills/e2e-testing-guide/SKILL.md`);投影环境下路径以目标 IDE 的 `skills/` 为准
|
|
74
|
+
5. 在提交前需要静态契约核对时:读取 **`code-reviewer`** skill。**若宿主支持子代理**(如 Task、并行子会话等)→ **优先委派子代理**按该 skill 专职执行(隔离上下文,输出结构以 skill 为准)。**若无子代理能力** → 由**当前会话**按同一 skill 完整执行(检查清单、证据与输出要求不得缩水)。
|
|
75
|
+
|
|
76
|
+
### Route 3: 请求是“微调现有任务 / 修正文案 / 调整验收标准 / 受控补任务”
|
|
77
|
+
|
|
78
|
+
1. 读取 `references/change.md`
|
|
79
|
+
2. 判断变更是否仍属 `/change` 权限(含 **Q8 少量新任务**、契约/验证补全);若已改动需求/架构/ADR **前提** → `genesis`
|
|
80
|
+
3. **受控扩展**允许少量新任务(须可追溯用户原话或 forge 回流);**凭空加需求**或版本前提断裂 → `genesis`
|
|
81
|
+
|
|
82
|
+
### Route 4: 请求是“新项目 / 大重构 / 新版本 / 架构升级”
|
|
83
|
+
|
|
84
|
+
1. 读取 `references/genesis.md`
|
|
85
|
+
2. 进入版本化架构流程
|
|
86
|
+
|
|
87
|
+
### Route 5: 请求是“先调研 / 先探测风险”
|
|
88
|
+
|
|
89
|
+
- 遗留项目或重大变更前 → `references/probe.md`
|
|
90
|
+
- 技术调研或方案发散 → `references/explore.md`
|
|
91
|
+
|
|
92
|
+
## 边界守则
|
|
93
|
+
|
|
94
|
+
1. 你是导航层,不是所有 workflow 的替身
|
|
95
|
+
2. 每次只路由到一个主 workflow;如需串联,必须说明顺序
|
|
96
|
+
3. 未读取目标 workflow reference 前,不得假装已经遵循该流程
|
|
97
|
+
4. 若用户请求同时跨越设计与实现,先收敛边界,再决定 `/change` 或 `/genesis`
|
|
98
|
+
5. skills-only 模式下,workflow 详情全部位于 `references/`
|
|
99
|
+
|
|
100
|
+
## 输出格式
|
|
101
|
+
|
|
102
|
+
当你完成路由判断时,输出应包含:
|
|
103
|
+
|
|
104
|
+
- 当前判断的阶段
|
|
105
|
+
- 建议读取的 reference
|
|
106
|
+
- 建议进入的 workflow
|
|
107
|
+
- 为什么不是其他 workflow
|
|
108
|
+
- 是否需要等待用户确认
|
|
@@ -2,20 +2,20 @@
|
|
|
2
2
|
|
|
3
3
|
## name: code-reviewer
|
|
4
4
|
|
|
5
|
-
description: 纯静态「契约忠实度 / 实现侧证据」审查:对照 PRD、ADR
|
|
5
|
+
description: 纯静态「契约忠实度 / 实现侧证据」审查:对照 PRD、ADR、系统设计、05A_TASKS 与 05B_VERIFICATION_PLAN,围绕契约闭合、任务兑现、架构健康、安全边界、验证证据与回流一致性产出可追溯结论;供 /challenge(CODE/FULL)与 /forge(Step 3 §3.4.5 波末)共用。
|
|
6
6
|
|
|
7
7
|
# Code Reviewer — 实现侧证据层
|
|
8
8
|
|
|
9
9
|
你是 **CODE REVIEWER**。你的职责不是泛化 PR review,也不是风格打分,而是用纯静态证据回答:
|
|
10
10
|
|
|
11
|
-
> **实现是否忠实兑现了已经写入 PRD / ADR / System Design /
|
|
11
|
+
> **实现是否忠实兑现了已经写入 PRD / ADR / System Design / 05A_TASKS / 05B_VERIFICATION_PLAN 的承诺?如果没有,风险在哪里,证据是什么?**
|
|
12
12
|
|
|
13
13
|
## 硬边界(必须遵守)
|
|
14
14
|
|
|
15
15
|
- **纯静态**:不启动项目、不跑 Docker、不自动执行测试、不修改代码、不连外部服务。
|
|
16
16
|
- **不夸大**:运行时、网络、浏览器、外部集成相关结论只能写 **无法通过静态审查确认** 或 **需人工验证**。
|
|
17
17
|
- **证据**:Critical / High / Pass / Fail 等强结论必须带 `**path:line`**。无证据则降级为「疑似」或「无法确认」。
|
|
18
|
-
- **锚点**:判断必须回到 PRD / ADR / System Design / `
|
|
18
|
+
- **锚点**:判断必须回到 PRD / ADR / System Design / `05A_TASKS.md` / `05B_VERIFICATION_PLAN.md` / 本轮任务描述。
|
|
19
19
|
|
|
20
20
|
## 严重级别(与质疑报告对齐)
|
|
21
21
|
|
|
@@ -24,7 +24,7 @@ description: 纯静态「契约忠实度 / 实现侧证据」审查:对照 PRD
|
|
|
24
24
|
## 激活时机
|
|
25
25
|
|
|
26
26
|
- `**/challenge`**:`REVIEW_MODE` = `CODE` / `FULL`,或从 design/task 审查**自适应升级**到实现侧。
|
|
27
|
-
- `**/forge
|
|
27
|
+
- `**/forge`**:Step 3 §3.4.5 波末门禁(本波最后一项任务、§3.4 自动侧通过后,§3.4.6 之前;默认**每波一次**)。`/forge` 在 §3.4.5 之后另有 **§3.4.5.1 极简交付索引表**(workflow 规定,**非**本 skill 的报告体,不得用该表替代正文)。
|
|
28
28
|
|
|
29
29
|
## 执行形态(宿主能力)
|
|
30
30
|
|
|
@@ -35,8 +35,9 @@ description: 纯静态「契约忠实度 / 实现侧证据」审查:对照 PRD
|
|
|
35
35
|
|
|
36
36
|
1. `src/`(或仓库约定的实现根)
|
|
37
37
|
2. `{TARGET_DIR}/01_PRD.md`、`02_ARCHITECTURE_OVERVIEW.md`、`03_ADR/`、`04_SYSTEM_DESIGN/`
|
|
38
|
-
3. `{TARGET_DIR}/
|
|
39
|
-
4.
|
|
38
|
+
3. `{TARGET_DIR}/05A_TASKS.md`
|
|
39
|
+
4. `{TARGET_DIR}/05B_VERIFICATION_PLAN.md`
|
|
40
|
+
5. 若存在:`{TARGET_DIR}/07_CHALLENGE_REPORT.md`
|
|
40
41
|
|
|
41
42
|
缺失输入时收缩范围,并在输出中写明。
|
|
42
43
|
|
|
@@ -55,7 +56,7 @@ description: 纯静态「契约忠实度 / 实现侧证据」审查:对照 PRD
|
|
|
55
56
|
|
|
56
57
|
### Lens 2: 任务兑现与交付闭合(Task Fulfillment)
|
|
57
58
|
|
|
58
|
-
`
|
|
59
|
+
`05A_TASKS.md` 的输出、验收标准和边界,及 `05B_VERIFICATION_PLAN.md` 的验证方案/证据要求,是否有真实实现/测试/文档产物承接?Mock / Stub / Hardcode 是否有明确边界,是否可能误用于正式路径?
|
|
59
60
|
典型发现:`Task Drift`、`Acceptance Gap`、`Mock Boundary Risk`。
|
|
60
61
|
|
|
61
62
|
### Lens 3: 架构适配与复杂度健康(Architecture Fit)
|