superpowers-subagent-dev

Use when executing implementation plans with independent tasks - coordinates task execution by dispatching subagents per task with verification checkpoints, adapted for OpenClaw's isolated session model

3,891 stars

Best use case

superpowers-subagent-dev is best used when you need a repeatable AI agent workflow instead of a one-off prompt.

Use when executing implementation plans with independent tasks - coordinates task execution by dispatching subagents per task with verification checkpoints, adapted for OpenClaw's isolated session model

Teams using superpowers-subagent-dev should expect a more consistent output, faster repeated execution, less prompt rewriting.

When to use this skill

  • You want a reusable workflow that can be run more than once with consistent structure.

When not to use this skill

  • You only need a quick one-off answer and do not need a reusable workflow.
  • You cannot install or maintain the underlying files, dependencies, or repository context.

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/superpowers-subagent-dev/SKILL.md --create-dirs "https://raw.githubusercontent.com/openclaw/skills/main/skills/axelhu/superpowers-subagent-dev/SKILL.md"

Manual Installation

  1. Download SKILL.md from GitHub
  2. Place it in .claude/skills/superpowers-subagent-dev/SKILL.md inside your project
  3. Restart your AI agent — it will auto-discover the skill

How superpowers-subagent-dev Compares

Feature / Agentsuperpowers-subagent-devStandard Approach
Platform SupportNot specifiedLimited / Varies
Context Awareness High Baseline
Installation ComplexityUnknownN/A

Frequently Asked Questions

What does this skill do?

Use when executing implementation plans with independent tasks - coordinates task execution by dispatching subagents per task with verification checkpoints, adapted for OpenClaw's isolated session model

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

Related Guides

SKILL.md Source

# Superpowers Subagent 开发(OpenClaw 适配版)

## 概述

通过为每个任务 dispatch 独立 subagent 来执行计划,每个任务后进行两阶段审查:先审查规格合规,再审查代码质量。

**为什么用 subagent:** 将任务委托给专门 agent,隔离上下文。通过精确构造指令和上下文,确保它们专注并成功完成。Subagent 不应继承 session 历史——精确构造它们需要的上下文。

**核心原则:** 每个任务新鲜 subagent + 两阶段审查(规格合规 → 代码质量)

## OpenClaw 适配说明

Superpowers 原版设计基于 Claude Code 的 Task/subagent 级联模型。OpenClaw 使用 `sessions_spawn` 创建独立 session,无法做到原生级联。

**OpenClaw 适配方案:**
- 主 agent 作为 controller,协调所有工作
- 用 `sessions_spawn(mode="run")` dispatch 一次性任务 subagent
- Subagent 结果通过 session 历史或文件系统传递
- 审查在主 session 内 inline 执行(或 spawn 独立审查 session)
- 复杂审查任务用 `sessions_spawn`

## 使用条件

```
有实现计划?
  → 任务大多独立?
    → 在本 session 内工作?
      → 用 subagent-dev(这个技能)
    → 并行独立 session?
      → 用 sessions_spawn 并行 dispatch
  → 手动执行或先 brainstorming
```

## 流程

### 每次任务

1. **读取计划,提取所有任务**
   - 读取计划文件一次
   - 提取所有任务及其完整文本和上下文
   - 创建任务列表

2. **Dispatch 实现者 subagent**
   ```
   使用 sessions_spawn:
   - mode: "run"(一次性任务)
   - task: 完整任务文本 + 上下文
   - cwd: 项目目录
   ```
   
3. **Subagent 提问处理**
   - 如果 subagent 提问 → 回答问题,提供上下文
   - 重新 dispatch 或继续

4. **Subagent 实现、测试、commit、自审**
   - 实现者完成工作
   - 运行测试
   - Commit
   - 自审

5. **规格合规审查(主 session inline 或 spawn)**
   - 检查实现是否满足计划中的规格
   - 发现问题 → 反馈给实现者修复 → 重新审查
   - 通过 → 继续

6. **代码质量审查**
   - 检查代码质量:DRY、命名、测试设计
   - 发现问题 → 修复
   - 通过 → 继续

7. **标记任务完成,继续下一个**

### 任务后

所有任务完成后:
- 做最终代码审查
- 调用 `superpowers-finishing-branch` 完成工作

## 模型选择

用能处理任务的最弱模型,节省成本增加速度:

| 任务类型 | 示例 | 模型 |
|---------|------|------|
| 机械实现任务 | 孤立函数、清晰规格、1-2 文件 | 快速便宜模型 |
| 集成和判断任务 | 多文件协调、模式匹配、调试 | 标准模型 |
| 架构设计和审查任务 | 需要设计判断或广泛代码库理解 | 最强模型 |

## 处理 Subagent 状态

Subagent 报告四种状态之一。适当处理:

**DONE:** 继续规格合规审查。

**DONE_WITH_CONCERNS:** 实现者完成但标记了疑虑。先读关注点再继续。如果关注点涉及正确性或范围,先解决再审查。

**NEEDS_CONTEXT:** 实现者需要未提供的信息。提供缺失上下文并重新 dispatch。

**BLOCKED:** 实现者无法完成任务。评估阻塞:
1. 如果是上下文问题,提供更多上下文并用相同模型重新 dispatch
2. 如果任务需要更多推理,用更强模型重新 dispatch
3. 如果任务太大,拆成更小的块
4. 如果计划本身有问题,上报给主人

**永远不要**忽略升级或强制相同模型重试而不做变更。

## Red Flags

**永远不要:**
- 在 main/master 分支上开始实现(没有主人明确同意)
- 跳过审查(规格合规或代码质量)
- 有未修复问题时继续
- 同时 dispatch 多个实现 subagent(冲突)
- 让 subagent 读计划文件(直接提供完整文本)
- 跳过场景设置上下文(subagent 需要理解任务在哪里_fit)
- 忽略 subagent 的提问(先回答再让他们继续)
- 接受"差不多"的规格合规(审查发现问题 = 没完成)
- 跳过审查循环(审查发现问题 = 实现者修复 = 重新审查)
- 让实现者自审替代实际审查(两者都需要)
- **在规格合规通过前开始代码质量审查**(顺序错误)
- 有公开问题时继续下一个任务

## 集成

**必需的工作流技能:**
- `superpowers-writing-plans` — 创建这个技能执行的计划
- `superpowers-finishing-branch` — 所有任务完成后的收尾
- `superpowers-tdd` — Subagent 每个任务遵循 TDD

**替代工作流:**
- `sessions_spawn` 并行 dispatch — 用于独立问题的并行调查
- 顺序执行 — 在主 session 内按批次执行

Related Skills

superpowers-executing-plans

3891
from openclaw/skills

Use when executing a written implementation plan in the current session with sequential task execution and review checkpoints - for when subagent-driven mode is not available

superpowers-writing-plans

3891
from openclaw/skills

Use when you have a spec or requirements for a multi-step task, before touching code - guides writing comprehensive implementation plans with bite-sized tasks, TDD, and DRY/YAGNI principles

superpowers-verification

3891
from openclaw/skills

Use when about to claim any work is complete, fixed, passing, or successful - requires running fresh verification commands and reading actual output before making any success claims; evidence before assertions always

superpowers-tdd

3891
from openclaw/skills

Use when implementing any feature or bugfix, before writing implementation code - enforces RED-GREEN-REFACTOR cycle: write failing test first, verify it fails, write minimal code, verify it passes, then refactor

superpowers-systematic-debugging

3891
from openclaw/skills

Use when encountering any bug, test failure, or unexpected behavior - enforces systematic four-phase debugging: root cause investigation, pattern analysis, hypothesis testing, and evidence-based fix verification

superpowers-parallel-agents

3891
from openclaw/skills

Use when facing 2 or more independent tasks that can be worked on without shared state - dispatches parallel subagents using sessions_spawn for concurrent investigation and execution, adapted for OpenClaw

superpowers-overview

3891
from openclaw/skills

Use when starting any development work or when unsure which superpowers development skill to use - provides entry point and navigation to the full superpowers skill suite for OpenClaw agents

superpowers-isolated-workspace

3891
from openclaw/skills

Use when starting feature work that needs isolation from current workspace - creates isolated git branches with clean setup and safety verification, adapted for OpenClaw environments

superpowers-finishing-branch

3891
from openclaw/skills

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - presents structured options for merge, PR, or cleanup; adapted for OpenClaw git workflow without worktrees

superpowers-brainstorming

3891
from openclaw/skills

Use before any creative work - creating features, building components, adding functionality, or modifying behavior - guides through exploration, questioning, design proposal, and spec documentation before any implementation

superpowers-requesting-code-review

3891
from openclaw/skills

Use when completing tasks, implementing major features, or before merging - dispatches code review subagent to catch issues before they cascade, adapted for OpenClaw sessions_spawn model

superpowers-receiving-code-review

3891
from openclaw/skills

Use when receiving code review feedback - requires technical verification before implementing suggestions, with reasoned pushback when feedback is technically questionable; no performative agreement