@wenaixi/dsh-superpower 6.3.0-dsh.9 → 6.3.1
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 +2 -2
- package/package.json +3 -3
- package/skills/superpower-brainstorming/SKILL.md +5 -5
- package/skills/superpower-executing-plans/SKILL.md +2 -2
- package/skills/superpower-subagent-driven-development/SKILL.md +2 -2
- package/skills/superpower-subagent-driven-development/re-review-prompt.md +1 -2
- package/skills/superpower-subagent-driven-development/task-reviewer-prompt.md +1 -2
- package/skills/superpower-using-superpowers/references/codex-tools.md +2 -2
- package/skills/superpower-using-superpowers/references/dsh-tools.md +1 -1
- package/skills/superpower-using-superpowers/references/hermes-tools.md +2 -2
- package/skills/superpower-writing-skills/SKILL.md +2 -2
package/README.md
CHANGED
|
@@ -59,7 +59,7 @@ dsh plugin --profile web remove @wenaixi/dsh-superpower
|
|
|
59
59
|
| `superpower-dispatching-parallel-agents` | 并行分发 |
|
|
60
60
|
| `superpower-test-driven-development` | RED-GREEN-REFACTOR |
|
|
61
61
|
| `superpower-systematic-debugging` / `superpower-verification-before-completion` | 调试闭环 |
|
|
62
|
-
| `superpower-requesting` / `superpower-receiving-code-review` | 评审 |
|
|
62
|
+
| `superpower-requesting-code-review` / `superpower-receiving-code-review` | 评审 |
|
|
63
63
|
| `superpower-finishing-a-development-branch` | 集成 |
|
|
64
64
|
| `superpower-writing-skills` | 写新技能 |
|
|
65
65
|
|
|
@@ -90,7 +90,7 @@ skills/ # 14 技能(中文化)
|
|
|
90
90
|
lib/ # 已提交,GitHub 直装零构建
|
|
91
91
|
```
|
|
92
92
|
|
|
93
|
-
|
|
93
|
+
版本:`v6.3.1` 起本仓库脱离上游独立演进(上游基准锁定 `obra/superpowers v6.3.0`),后续变更以正式版本线发布;`tag v*` 触发发布,`push` 仅跑 CI。详见 `CHANGELOG.md`。
|
|
94
94
|
|
|
95
95
|
## 常见问题
|
|
96
96
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@wenaixi/dsh-superpower",
|
|
3
|
-
"version": "6.3.
|
|
3
|
+
"version": "6.3.1",
|
|
4
4
|
"description": "obra/superpowers 的 DSH 移植版 — 将完整的多智能体开发方法论以原生 DSH 技能开箱即用",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "lib/superpowers.js",
|
|
@@ -30,12 +30,12 @@
|
|
|
30
30
|
},
|
|
31
31
|
"peerDependencies": {
|
|
32
32
|
"@deepseek-ai/cordis": "^4.0.1",
|
|
33
|
-
"@deepseek-ai/dsh-skill": "
|
|
33
|
+
"@deepseek-ai/dsh-skill": ">=0.0.1-rc.1 <0.1.0 || >=0.1.0-rc.1 <0.2.0-0",
|
|
34
34
|
"@deepseek-ai/schemastery": "^3.18.1"
|
|
35
35
|
},
|
|
36
36
|
"devDependencies": {
|
|
37
37
|
"@deepseek-ai/cordis": "^4.0.1",
|
|
38
|
-
"@deepseek-ai/dsh-skill": "
|
|
38
|
+
"@deepseek-ai/dsh-skill": "0.1.1-rc.2",
|
|
39
39
|
"@deepseek-ai/schemastery": "^3.18.1",
|
|
40
40
|
"typescript": "^5.6.0",
|
|
41
41
|
"@types/node": "^22.0.0"
|
|
@@ -19,7 +19,7 @@ description: "创意工作前必用:创建功能、构建组件、新增能力
|
|
|
19
19
|
|
|
20
20
|
- **Spike 探索** — 可行性问题(“能否……”“是否可能……”“快速粗糙实现即可”),产出是答案,而非可保留的代码。用 2-3 句话说明问题及验证方式,获得认可后,以满足正确性前提下的最低成本去验证。无需设计文档与规格文件。以建议形式汇报发现,任何为验证而构建的代码均标注为“一次性/可丢弃”。
|
|
21
21
|
- **Bounded 受限** — 对本仓库已有代码的明确范围变更:新增开关、小型接口、单文件修复。仅了解应用类型是不够的——“受限”意味着你要改的流程已在仓库中可查可读。若没有可改的既有流程,则不属于受限任务。提出关键的澄清问题,在对话中直接呈现简短设计(几句话到几段短文),然后停止。仅在协作方对该设计明确说“同意”后方可开始实现——受限任务的审批与架构级任务一样严格。无需规格文件与实现计划文档。
|
|
22
|
-
- **Architectural 架构** — 新项目、新子系统,或会重塑组件间协作方式、改变他人依赖接口的变更。遵循完整流程:提问、提供方案、分节设计、编写规格文档,然后进入 writing-plans
|
|
22
|
+
- **Architectural 架构** — 新项目、新子系统,或会重塑组件间协作方式、改变他人依赖接口的变更。遵循完整流程:提问、提供方案、分节设计、编写规格文档,然后进入 superpower-writing-plans 技能。
|
|
23
23
|
|
|
24
24
|
当在两条路径间犹豫时,选择更重的那一条。棘轮单向生效:任务中途发现隐藏复杂度即升级路径——立即停下、说明情况、提升等级。任务中途不允许降级。
|
|
25
25
|
|
|
@@ -66,7 +66,7 @@ description: "创意工作前必用:创建功能、构建组件、新增能力
|
|
|
66
66
|
6. **编写设计文档** — 保存至 `docs/superpowers/specs/YYYY-MM-DD-<topic>-design.md` 并提交
|
|
67
67
|
7. **规格自检** — 快速行内检查占位符、矛盾、歧义与范围(见下文)
|
|
68
68
|
8. **用户复核书面规格** — 请用户在继续前复核规格文件
|
|
69
|
-
9. **转入实现** — 调用 writing-plans
|
|
69
|
+
9. **转入实现** — 调用 superpower-writing-plans 技能创建实现计划
|
|
70
70
|
|
|
71
71
|
## 流程图
|
|
72
72
|
|
|
@@ -87,7 +87,7 @@ digraph brainstorming {
|
|
|
87
87
|
"编写设计文档" [shape=box];
|
|
88
88
|
"规格自检\n(行内修复)" [shape=box];
|
|
89
89
|
"用户是否复核规格?" [shape=diamond];
|
|
90
|
-
"调用 writing-plans
|
|
90
|
+
"调用 superpower-writing-plans 技能" [shape=doublecircle];
|
|
91
91
|
"发现隐藏复杂度?升级路径" [shape=box];
|
|
92
92
|
|
|
93
93
|
"分类:spike 探索 / bounded 受限 / architectural 架构" -> "呈现问题与验证计划(2-3 句话)" [label="spike 探索"];
|
|
@@ -108,7 +108,7 @@ digraph brainstorming {
|
|
|
108
108
|
"编写设计文档" -> "规格自检\n(行内修复)";
|
|
109
109
|
"规格自检\n(行内修复)" -> "用户是否复核规格?";
|
|
110
110
|
"用户是否复核规格?" -> "编写设计文档" [label="需修改"];
|
|
111
|
-
"用户是否复核规格?" -> "调用 writing-plans
|
|
111
|
+
"用户是否复核规格?" -> "调用 superpower-writing-plans 技能" [label="已批准"];
|
|
112
112
|
}
|
|
113
113
|
```
|
|
114
114
|
|
|
@@ -184,7 +184,7 @@ digraph brainstorming {
|
|
|
184
184
|
|
|
185
185
|
**实现:**
|
|
186
186
|
|
|
187
|
-
- 调用 writing-plans
|
|
187
|
+
- 调用 superpower-writing-plans 技能创建详细的实现计划
|
|
188
188
|
- 不要调用任何其他 skill。下一步仅为 writing-plans。
|
|
189
189
|
|
|
190
190
|
## 可视化协作
|
|
@@ -9,7 +9,7 @@ description: "适用于已有详细实现计划的场景,在隔离会话中逐
|
|
|
9
9
|
|
|
10
10
|
加载计划、批判性复审、执行全部任务、完成后汇报。
|
|
11
11
|
|
|
12
|
-
**开始时声明:** "
|
|
12
|
+
**开始时声明:** "我将使用 executing-plans 技能来执行本计划。"
|
|
13
13
|
|
|
14
14
|
**提示:** 告知你的协作伙伴,Superpowers 在可使用 subagent 时效果更好(Claude Code、Codex CLI、Codex App、Copilot CLI 和 Gemini CLI 均符合条件;各平台工具说明见 `../using-superpowers/references/`)。如可使用 subagent,请使用 superpower-subagent-driven-development 替代本 skill。
|
|
15
15
|
|
|
@@ -33,7 +33,7 @@ description: "适用于已有详细实现计划的场景,在隔离会话中逐
|
|
|
33
33
|
### 步骤 3:完成开发
|
|
34
34
|
|
|
35
35
|
所有任务完成并校验通过后:
|
|
36
|
-
- 声明:"
|
|
36
|
+
- 声明:"我将使用 finishing-a-development-branch 技能来完成收尾工作。"
|
|
37
37
|
- **必选子 skill:** 使用 superpower-finishing-a-development-branch
|
|
38
38
|
- 按该 skill 流程校验测试、提供选项并执行所选方案
|
|
39
39
|
|
|
@@ -27,7 +27,7 @@ digraph when_to_use {
|
|
|
27
27
|
"任务是否大多相互独立?" [shape=diamond];
|
|
28
28
|
"是否留在当前会话?" [shape=diamond];
|
|
29
29
|
"subagent-driven-development" [shape=box];
|
|
30
|
-
"executing-plans" [shape=box];
|
|
30
|
+
"superpower-executing-plans" [shape=box];
|
|
31
31
|
"手动执行或先行头脑风暴" [shape=box];
|
|
32
32
|
|
|
33
33
|
"是否有实现计划?" -> "任务是否大多相互独立?" [label="是"];
|
|
@@ -35,7 +35,7 @@ digraph when_to_use {
|
|
|
35
35
|
"任务是否大多相互独立?" -> "是否留在当前会话?" [label="是"];
|
|
36
36
|
"任务是否大多相互独立?" -> "手动执行或先行头脑风暴" [label="否 - 高度耦合"];
|
|
37
37
|
"是否留在当前会话?" -> "subagent-driven-development" [label="是"];
|
|
38
|
-
"是否留在当前会话?" -> "executing-plans" [label="否 - 并行会话"];
|
|
38
|
+
"是否留在当前会话?" -> "superpower-executing-plans" [label="否 - 并行会话"];
|
|
39
39
|
}
|
|
40
40
|
```
|
|
41
41
|
|
|
@@ -7,8 +7,7 @@
|
|
|
7
7
|
```
|
|
8
8
|
Subagent (general-purpose):
|
|
9
9
|
description: "Re-review Task N fix round R"
|
|
10
|
-
model: [MODEL —
|
|
11
|
-
model silently inherits the session's most expensive one]
|
|
10
|
+
model: [MODEL — 必填:按 SKILL.md 的模型选择规则选择;若省略则静默继承会话中最昂贵的模型]
|
|
12
11
|
prompt: |
|
|
13
12
|
你正在复审单个任务的某一轮修复。上一轮评审已产生若干问题,执行人已尝试修复。你的职责仅是对每一条问题给出裁决,并检查修复 diff——除此之外不做其他事。
|
|
14
13
|
|
|
@@ -7,8 +7,7 @@
|
|
|
7
7
|
```
|
|
8
8
|
Subagent (general-purpose):
|
|
9
9
|
description: "Review Task N (spec + quality)"
|
|
10
|
-
model: [MODEL —
|
|
11
|
-
model silently inherits the session's most expensive one]
|
|
10
|
+
model: [MODEL — 必填:按 SKILL.md 的模型选择规则选择;若省略则静默继承会话中最昂贵的模型]
|
|
12
11
|
prompt: |
|
|
13
12
|
你正在评审单个任务的实现:先判断是否符合需求,再判断是否构建良好。这是一个任务级别的门禁,而非合并评审——所有任务完成后的全分支广度评审会另行进行。
|
|
14
13
|
|
|
@@ -8,7 +8,7 @@ multi_agent = true
|
|
|
8
8
|
```
|
|
9
9
|
|
|
10
10
|
This enables the multi-agent tools that skills like
|
|
11
|
-
`dispatching-parallel-agents` and `subagent-driven-development` use.
|
|
11
|
+
`superpower-dispatching-parallel-agents` and `superpower-subagent-driven-development` use.
|
|
12
12
|
Which tools you get depends on the multi-agent version your model
|
|
13
13
|
preset selects (current presets run V2; older ones run V1). Trust your
|
|
14
14
|
actual tool list over any table — including this one — when they
|
|
@@ -92,7 +92,7 @@ BRANCH=$(git branch --show-current)
|
|
|
92
92
|
- `GIT_DIR != GIT_COMMON` → already in a linked worktree (skip creation)
|
|
93
93
|
- `BRANCH` empty → detached HEAD (cannot branch/push/PR from sandbox)
|
|
94
94
|
|
|
95
|
-
See `using-git-worktrees` Step 0 and `finishing-a-development-branch`
|
|
95
|
+
See `superpower-using-git-worktrees` Step 0 and `superpower-finishing-a-development-branch`
|
|
96
96
|
Step 1 for how each skill uses these signals.
|
|
97
97
|
|
|
98
98
|
## Codex App Finishing
|
|
@@ -25,7 +25,7 @@
|
|
|
25
25
|
| `goal` / `ralph` | 长周期目标的创建、续跑、阻塞判定 |
|
|
26
26
|
| `workflow` 脚本 | 需要编排数十个 subagent 的审计/迁移/批量改写 |
|
|
27
27
|
| `jobs` | 托管后台任务的查询与终止 |
|
|
28
|
-
| `plan-mode` | 只读规划锁定,与 `writing-plans` 的细粒度切片互补 |
|
|
28
|
+
| `plan-mode` | 只读规划锁定,与 `superpower-writing-plans` 的细粒度切片互补 |
|
|
29
29
|
| `cordis` 动态插件 | 会话内临时扩展能力(非本移植包范畴) |
|
|
30
30
|
|
|
31
31
|
## 约束与习惯
|
|
@@ -28,8 +28,8 @@ Hermes Agent has a `skills` toolset with `skill_view` and `skills_list` tools.
|
|
|
28
28
|
To invoke a superpowers skill, use:
|
|
29
29
|
|
|
30
30
|
```
|
|
31
|
-
skill_view("brainstorming")
|
|
32
|
-
skill_view("test-driven-development")
|
|
31
|
+
skill_view("superpower-brainstorming")
|
|
32
|
+
skill_view("superpower-test-driven-development")
|
|
33
33
|
```
|
|
34
34
|
|
|
35
35
|
If `skill_view` cannot find a superpowers skill (it may not appear in the catalog
|
|
@@ -151,9 +151,9 @@ description: Use when [具体的触发条件与症状]
|
|
|
151
151
|
|
|
152
152
|
description 应仅描述触发条件,不要概括技能的流程或工作流。
|
|
153
153
|
|
|
154
|
-
**为什么这很重要:** 测试发现,当 description 概括了技能的工作流时,智能体可能会直接沿用 description
|
|
154
|
+
**为什么这很重要:** 测试发现,当 description 概括了技能的工作流时,智能体可能会直接沿用 description 的描述,而不再阅读完整的技能内容。一个写着「code review between tasks」的 description 曾导致智能体只做了一次审查,尽管技能中的流程图明确展示了两次审查(先合规性审查,再代码质量审查)。
|
|
155
155
|
|
|
156
|
-
当 description
|
|
156
|
+
当 description 被改为仅保留「Use when executing implementation plans with independent tasks」(不含工作流概括)后,智能体才正确读取了流程图,并执行了两阶段审查流程。
|
|
157
157
|
|
|
158
158
|
**陷阱:** 概括工作流的 description 会成为智能体走捷径的入口,技能正文反而被跳过。
|
|
159
159
|
|