You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
feat: v2.2.0 — update all 13 agents to VS Code official tools, add structured output paths
- Replace generic tool names (search/read/edit/execute) with VS Code official
built-in tool names (codebase, textSearch, fileSearch, readFile, etc.)
- Add Playwright MCP tools (playwright/*) to 6 browser-capable agents:
ui-design, execute, tdd, debug, verify, review
- Migrate output paths from docs/plans/ and docs/ux/ to .github/superpower/
(brainstorm/, ux/, context/, plan/)
- Add inter-agent artifact discovery: plan reads from brainstorm/ and context/,
execute reads from plan/
- Update both English and Chinese READMEs with output path conventions and
optional Playwright MCP configuration instructions
- Bump version to 2.2.0
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
**9 structured AI workflow agents for GitHub Copilot Chat**
3
+
[中文文档](readme-zh.md)
4
+
5
+
**13 structured AI workflow agents for GitHub Copilot Chat**
4
6
5
7
Superpower Copilot brings disciplined, process-driven development workflows to GitHub Copilot through VS Code's [Custom Agents](https://code.visualstudio.com/docs/copilot/chat/chat-agents) feature. Each agent enforces a specific development practice — from brainstorming and planning to TDD, debugging, and code review — so you can work with the same rigor that professional engineering teams demand.
6
8
@@ -13,54 +15,76 @@ GitHub Copilot is powerful, but it lacks opinionated development workflows. With
13
15
- Debugging by guessing instead of systematic root-cause analysis
14
16
- Claiming work is "done" without verification
15
17
16
-
Superpower Copilot solves this by providing 9 agents, each with a rigorous system prompt (200-300 lines) that enforces best practices. The agents are inspired by the [Superpowers](https://github.com/cline/superpowers) skill framework and adapted for GitHub Copilot's agent architecture.
18
+
Superpower Copilot solves this by providing 13 agents, each with a rigorous system prompt (200-300 lines) that enforces best practices. The agents are inspired by the [Superpowers](https://github.com/cline/superpowers) skill framework and the [awesome-copilot](https://github.com/github/awesome-copilot) community, adapted for GitHub Copilot's agent architecture.
Open Copilot Chat (Ctrl+Shift+I / Cmd+Shift+I) and type:
96
+
1. Open Copilot Chat panel: `Ctrl+Alt+I` (Windows/Linux) / `Cmd+Alt+I` (Mac)
97
+
2. Click the agent dropdown at the top of the chat panel
98
+
3. Select a `superpower-*` agent from the list
99
+
4. Type your request and send
73
100
74
-
```
75
-
@superpower-brainstorm I need to add user authentication to my Express app
76
-
```
101
+
**Examples:**
77
102
78
-
```
79
-
@superpower-tdd Implement a shopping cart with add/remove/total
80
-
```
103
+
| Agent | Example prompt |
104
+
|-------|---------------|
105
+
|`superpower-brainstorm`| "I need to add user authentication to my Express app" |
106
+
|`superpower-tdd`| "Implement a shopping cart with add/remove/total" |
107
+
|`superpower-debug`| "Why is the checkout endpoint returning 500?" |
108
+
|`superpower-think`| "Should we use microservices or a monolith?" |
109
+
|`superpower-ui-design`| "Design the onboarding flow for our mobile app" |
110
+
|`superpower-context`| "I need to refactor the auth module across multiple files" |
111
+
|`superpower-mentor`| "Help me understand how the middleware chain works" |
81
112
82
-
```
83
-
@superpower-debug Why is the checkout endpoint returning 500?
84
-
```
85
-
86
-
The agent will guide you through its process step by step. Each agent enforces gates — for example, `@superpower-debug` will not let you propose fixes until root cause investigation is complete.
113
+
The agent will guide you through its process step by step. Each agent enforces gates — for example, `superpower-debug` will not let you propose fixes until root cause investigation is complete.
87
114
88
115
## What Makes These Agents Different
89
116
@@ -95,31 +122,72 @@ The agent will guide you through its process step by step. Each agent enforces g
95
122
96
123
**Workflow handoffs.** Agents include handoff buttons that connect to the next logical step in the chain.
97
124
125
+
**Inter-agent awareness.** Each agent knows about the others and includes Related Agents references in its description, so Copilot can suggest the right agent for the next step.
126
+
98
127
## Best Practices
99
128
100
129
-**Start with brainstorm** for any non-trivial feature. Spending 10 minutes on design saves hours of rework.
101
-
-**Use TDD for all new code.** The `@superpower-tdd` agent enforces Red-Green-Refactor. If you didn't see the test fail, you don't know it tests the right thing.
102
-
-**Verify before claiming done.** The `@superpower-verify` agent requires running commands and showing output. No "should work" or "looks correct".
103
-
-**Debug systematically.** When something breaks, resist the urge to guess. Use `@superpower-debug` to trace root cause first.
104
-
-**Review your own work.** Use `@superpower-review` before creating a PR. It catches issues you missed.
130
+
-**Map context first** for multi-file changes. Use `superpower-context` before `superpower-plan`.
131
+
-**Do UX research** before building UI. Use `superpower-ui-design` to understand users first.
132
+
-**Use TDD for all new code.** The `superpower-tdd` agent enforces Red-Green-Refactor. If you didn't see the test fail, you don't know it tests the right thing.
133
+
-**Verify before claiming done.** The `superpower-verify` agent requires running commands and showing output. No "should work" or "looks correct".
134
+
-**Debug systematically.** When something breaks, resist the urge to guess. Use `superpower-debug` to trace root cause first.
135
+
-**Challenge your thinking.** Use `superpower-think` when a decision feels too easy — it will expose blind spots.
136
+
-**Review your own work.** Use `superpower-review` before creating a PR. It catches issues you missed.
105
137
106
138
## How It Works
107
139
108
140
On activation, the extension:
109
141
110
-
1. Copies 9`.agent.md` files from the extension bundle to `~/.superpower-copilot/agents/`
142
+
1. Copies 13`.agent.md` files from the extension bundle to `~/.superpower-copilot/agents/`
111
143
2. Registers the path `~/.superpower-copilot/agents` in VS Code's `chat.agentFilesLocations` setting (user-level)
112
144
3. VS Code discovers the agents and adds them to the Copilot Chat dropdown
113
145
114
146
On deactivation, the extension cleans up the copied files and removes the setting entry.
115
147
148
+
### Agent Output Artifacts
149
+
150
+
Agents that produce documents (brainstorm, ui-design, context, plan) save their output to `.github/superpower/` in your workspace:
151
+
152
+
```
153
+
.github/superpower/
154
+
brainstorm/YYYY-MM-DD-<topic>-design.md
155
+
ux/[feature]-jtbd.md
156
+
ux/[feature]-journey.md
157
+
ux/[feature]-flow.md
158
+
context/YYYY-MM-DD-<topic>-context-map.md
159
+
plan/YYYY-MM-DD-<topic>-plan.md
160
+
```
161
+
162
+
Agents discover each other's artifacts automatically:
163
+
-`superpower-plan` reads from `brainstorm/` and `context/`
164
+
-`superpower-execute` reads from `plan/`
165
+
-`superpower-verify` and `superpower-review` read from `plan/`
166
+
167
+
### Playwright MCP (Optional)
168
+
169
+
For agents that interact with browsers (ui-design, debug, verify, execute, tdd, review), you can optionally configure the [Playwright MCP](https://github.com/anthropics/playwright-mcp) server. Add to your `.vscode/settings.json`:
170
+
171
+
```json
172
+
{
173
+
"mcpServers": {
174
+
"playwright": {
175
+
"command": "npx",
176
+
"args": ["@anthropic-ai/playwright-mcp@latest"]
177
+
}
178
+
}
179
+
}
180
+
```
181
+
182
+
This enables browser automation tools (`playwright/*`) for UI testing, visual verification, and frontend debugging. If not configured, agents work normally without browser capabilities.
183
+
116
184
## Language Support
117
185
118
186
All agents support both English and Simplified Chinese.
119
187
120
188
## Credits
121
189
122
-
Inspired by the [Superpowers](https://github.com/cline/superpowers) skill framework. Adapted from shell-based skills to VS Code Custom Agents with native Copilot integration (tools, handoffs, agent references).
190
+
Inspired by the [Superpowers](https://github.com/cline/superpowers) skill framework and the [awesome-copilot](https://github.com/github/awesome-copilot) community collection. Adapted for GitHub Copilot's Custom Agents with native integration (tools, handoffs, agent references).
0 commit comments