@su-record/vibe 1.1.10 → 1.2.1
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/.claude/commands/vibe.run.md +239 -30
- package/.claude/settings.local.json +2 -1
- package/CLAUDE.md +17 -0
- package/README.md +65 -11
- package/package.json +4 -2
- package/templates/hooks-template.json +15 -4
|
@@ -12,8 +12,126 @@ Execute implementation based on SPEC (Implementation Agent with Multi-Model Orch
|
|
|
12
12
|
```
|
|
13
13
|
/vibe.run "feature-name" # Full implementation
|
|
14
14
|
/vibe.run "feature-name" --phase 1 # Specific Phase only
|
|
15
|
+
/vibe.run "feature-name" ultrawork # ULTRAWORK mode (recommended)
|
|
16
|
+
/vibe.run "feature-name" ulw # Short alias for ultrawork
|
|
15
17
|
```
|
|
16
18
|
|
|
19
|
+
---
|
|
20
|
+
|
|
21
|
+
## **ULTRAWORK Mode** (ulw)
|
|
22
|
+
|
|
23
|
+
> Include `ultrawork` or `ulw` in your command to activate **maximum performance mode**.
|
|
24
|
+
|
|
25
|
+
### What ULTRAWORK Enables
|
|
26
|
+
|
|
27
|
+
When you include `ultrawork` (or `ulw`), ALL of these activate automatically:
|
|
28
|
+
|
|
29
|
+
| Feature | Description |
|
|
30
|
+
|---------|-------------|
|
|
31
|
+
| **Parallel Exploration** | 3+ Task(haiku) agents run simultaneously |
|
|
32
|
+
| **Boulder Loop** | Auto-continues until ALL phases complete |
|
|
33
|
+
| **Context Compression** | Aggressive auto-save at 70%+ context |
|
|
34
|
+
| **No Pause** | Doesn't wait for confirmation between phases |
|
|
35
|
+
| **External LLMs** | Auto-consults GPT/Gemini if enabled |
|
|
36
|
+
| **Error Recovery** | Auto-retries on failure (up to 3 times) |
|
|
37
|
+
|
|
38
|
+
### Boulder Loop (Inspired by Sisyphus)
|
|
39
|
+
|
|
40
|
+
Like Sisyphus rolling the boulder, ULTRAWORK **keeps going until done**:
|
|
41
|
+
|
|
42
|
+
```
|
|
43
|
+
┌─────────────────────────────────────────────────────────────────┐
|
|
44
|
+
│ BOULDER LOOP (ultrawork) │
|
|
45
|
+
│ │
|
|
46
|
+
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
|
|
47
|
+
│ │ Phase 1 │───→│ Phase 2 │───→│ Phase 3 │───→│ Phase N │ │
|
|
48
|
+
│ └──────────┘ └──────────┘ └──────────┘ └──────────┘ │
|
|
49
|
+
│ │ │ │ │ │
|
|
50
|
+
│ ↓ ↓ ↓ ↓ │
|
|
51
|
+
│ [Parallel] [Parallel] [Parallel] [Parallel] │
|
|
52
|
+
│ [Implement] [Implement] [Implement] [Implement] │
|
|
53
|
+
│ [Test] [Test] [Test] [Test] │
|
|
54
|
+
│ │ │ │ │ │
|
|
55
|
+
│ └───────────────┴───────────────┴───────────────┘ │
|
|
56
|
+
│ │ │
|
|
57
|
+
│ ↓ │
|
|
58
|
+
│ ┌──────────────┐ │
|
|
59
|
+
│ │ ALL DONE? │ │
|
|
60
|
+
│ └──────────────┘ │
|
|
61
|
+
│ │ │ │
|
|
62
|
+
│ NO YES │
|
|
63
|
+
│ │ │ │
|
|
64
|
+
│ ↓ ↓ │
|
|
65
|
+
│ [Continue] [🎉 Complete!] │
|
|
66
|
+
│ │
|
|
67
|
+
│ NO STOPPING until acceptance criteria met or error limit hit │
|
|
68
|
+
└─────────────────────────────────────────────────────────────────┘
|
|
69
|
+
```
|
|
70
|
+
|
|
71
|
+
### ULTRAWORK Example
|
|
72
|
+
|
|
73
|
+
```
|
|
74
|
+
User: /vibe.run "brick-game" ultrawork
|
|
75
|
+
|
|
76
|
+
Claude:
|
|
77
|
+
🚀 ULTRAWORK MODE ACTIVATED
|
|
78
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
79
|
+
|
|
80
|
+
📄 SPEC: .vibe/specs/brick-game.md
|
|
81
|
+
🎯 4 Phases detected
|
|
82
|
+
⚡ Boulder Loop: ENABLED (will continue until all phases complete)
|
|
83
|
+
🔄 Auto-retry: ON (max 3 per phase)
|
|
84
|
+
💾 Context compression: AGGRESSIVE
|
|
85
|
+
|
|
86
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
87
|
+
🏔️ BOULDER ROLLING... Phase 1/4
|
|
88
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
89
|
+
|
|
90
|
+
⚡ [PARALLEL] Launching 3 exploration agents...
|
|
91
|
+
✅ Exploration complete (7.2s)
|
|
92
|
+
🔨 Implementing...
|
|
93
|
+
✅ Phase 1 complete
|
|
94
|
+
|
|
95
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
96
|
+
🏔️ BOULDER ROLLING... Phase 2/4
|
|
97
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
98
|
+
|
|
99
|
+
⚡ [PARALLEL] Launching 3 exploration agents...
|
|
100
|
+
✅ Exploration complete (6.8s)
|
|
101
|
+
🔨 Implementing...
|
|
102
|
+
❌ Test failed: collision detection
|
|
103
|
+
🔄 Auto-retry 1/3...
|
|
104
|
+
🔨 Fixing...
|
|
105
|
+
✅ Phase 2 complete
|
|
106
|
+
|
|
107
|
+
[...continues automatically...]
|
|
108
|
+
|
|
109
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
110
|
+
🎉 BOULDER REACHED THE TOP!
|
|
111
|
+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
|
112
|
+
|
|
113
|
+
✅ All 4 phases complete
|
|
114
|
+
✅ All acceptance criteria passed
|
|
115
|
+
✅ Build succeeded
|
|
116
|
+
✅ Tests passed
|
|
117
|
+
|
|
118
|
+
⏱️ Total: 8m 24s
|
|
119
|
+
📊 Retries: 2
|
|
120
|
+
💾 Context saved: 3 checkpoints
|
|
121
|
+
```
|
|
122
|
+
|
|
123
|
+
### Normal vs ULTRAWORK Comparison
|
|
124
|
+
|
|
125
|
+
| Aspect | Normal | ULTRAWORK |
|
|
126
|
+
|--------|--------|-----------|
|
|
127
|
+
| Phase transition | May pause | Auto-continues |
|
|
128
|
+
| On error | Reports and stops | Auto-retries (3x) |
|
|
129
|
+
| Context 70%+ | Warning only | Auto-compress + save |
|
|
130
|
+
| Exploration | Sequential possible | FORCED parallel |
|
|
131
|
+
| Completion | Phase-by-phase | Until ALL done |
|
|
132
|
+
|
|
133
|
+
---
|
|
134
|
+
|
|
17
135
|
## Rules Reference
|
|
18
136
|
|
|
19
137
|
**Must follow `.vibe/rules/`:**
|
|
@@ -135,48 +253,110 @@ Check `.vibe/config.json`:
|
|
|
135
253
|
- Check BDD Scenarios
|
|
136
254
|
- Use as test cases
|
|
137
255
|
|
|
138
|
-
### 3. Phase-by-Phase Implementation
|
|
256
|
+
### 3. Phase-by-Phase Implementation
|
|
257
|
+
|
|
258
|
+
Follow `<task>` section Phase order.
|
|
259
|
+
|
|
260
|
+
---
|
|
261
|
+
|
|
262
|
+
## **CRITICAL: Parallel Sub-Agent Execution**
|
|
139
263
|
|
|
140
|
-
|
|
264
|
+
> **MUST USE PARALLEL TASK CALLS** - This is REQUIRED, not optional.
|
|
265
|
+
> Sequential execution when parallel is possible = VIOLATION of this workflow.
|
|
266
|
+
|
|
267
|
+
### Mandatory Parallel Exploration (Phase Start)
|
|
268
|
+
|
|
269
|
+
**BEFORE any implementation, you MUST launch these Task calls IN PARALLEL (single message, multiple tool calls):**
|
|
141
270
|
|
|
142
271
|
```
|
|
143
|
-
|
|
144
|
-
|
|
145
|
-
|
|
146
|
-
|
|
272
|
+
┌─────────────────────────────────────────────────────────────────┐
|
|
273
|
+
│ STEP 1: PARALLEL EXPLORATION (REQUIRED) │
|
|
274
|
+
│ │
|
|
275
|
+
│ Launch ALL of these in ONE message: │
|
|
276
|
+
│ │
|
|
277
|
+
│ Task(haiku) ─┬─→ "Analyze related files in <context>" │
|
|
278
|
+
│ │ │
|
|
279
|
+
│ Task(haiku) ─┼─→ "Check dependencies and imports" │
|
|
280
|
+
│ │ │
|
|
281
|
+
│ Task(haiku) ─┴─→ "Find existing patterns and conventions" │
|
|
282
|
+
│ │
|
|
283
|
+
│ [If GPT enabled] + MCP(vibe-gpt): Architecture review │
|
|
284
|
+
│ [If Gemini enabled] + MCP(vibe-gemini): UI/UX consultation │
|
|
285
|
+
└─────────────────────────────────────────────────────────────────┘
|
|
286
|
+
│
|
|
287
|
+
↓ (wait for all to complete)
|
|
288
|
+
┌─────────────────────────────────────────────────────────────────┐
|
|
289
|
+
│ STEP 2: SYNTHESIZE (Opus) │
|
|
290
|
+
│ - Review all exploration results │
|
|
291
|
+
│ - Decide implementation approach │
|
|
292
|
+
│ - Identify files to modify/create │
|
|
293
|
+
└─────────────────────────────────────────────────────────────────┘
|
|
294
|
+
│
|
|
295
|
+
↓
|
|
296
|
+
┌─────────────────────────────────────────────────────────────────┐
|
|
297
|
+
│ STEP 3: IMPLEMENT (Task sonnet) │
|
|
298
|
+
│ - Execute implementation with full context │
|
|
299
|
+
└─────────────────────────────────────────────────────────────────┘
|
|
300
|
+
│
|
|
301
|
+
↓
|
|
302
|
+
┌─────────────────────────────────────────────────────────────────┐
|
|
303
|
+
│ STEP 4: TEST (Task haiku) │
|
|
304
|
+
│ - Write tests for implemented code │
|
|
305
|
+
└─────────────────────────────────────────────────────────────────┘
|
|
306
|
+
```
|
|
307
|
+
|
|
308
|
+
### Parallel Task Call Pattern (MUST FOLLOW)
|
|
309
|
+
|
|
310
|
+
**Correct - Single message with multiple parallel Tasks:**
|
|
311
|
+
```
|
|
312
|
+
<message>
|
|
313
|
+
Task(haiku, "Analyze src/components/ for existing patterns")
|
|
314
|
+
Task(haiku, "Check package.json dependencies")
|
|
315
|
+
Task(haiku, "Find usage of similar features in codebase")
|
|
316
|
+
</message>
|
|
317
|
+
→ All 3 run simultaneously, ~3x faster
|
|
318
|
+
```
|
|
319
|
+
|
|
320
|
+
**WRONG - Sequential calls (DO NOT DO THIS):**
|
|
321
|
+
```
|
|
322
|
+
<message>Task(haiku, "Analyze...")</message>
|
|
323
|
+
<message>Task(haiku, "Check...")</message>
|
|
324
|
+
<message>Task(haiku, "Find...")</message>
|
|
325
|
+
→ 3x slower, wastes time
|
|
326
|
+
```
|
|
327
|
+
|
|
328
|
+
### Why Parallel Matters
|
|
329
|
+
|
|
330
|
+
| Approach | Time | Cache Benefit |
|
|
331
|
+
|----------|------|---------------|
|
|
332
|
+
| Sequential (3 Tasks) | ~30s | Cache cold on each |
|
|
333
|
+
| **Parallel (3 Tasks)** | **~10s** | **Cache warmed once, shared** |
|
|
334
|
+
|
|
335
|
+
hi-ai ProjectCache (LRU) caches ts-morph parsing results. Parallel calls share the warmed cache.
|
|
336
|
+
|
|
337
|
+
### Phase Execution Flow
|
|
338
|
+
|
|
339
|
+
```
|
|
340
|
+
Phase N Start
|
|
147
341
|
│
|
|
148
|
-
├─→ [
|
|
149
|
-
│
|
|
342
|
+
├─→ [PARALLEL] Task(haiku) × 2-3: Exploration
|
|
343
|
+
│ - Related code analysis
|
|
344
|
+
│ - Dependency check
|
|
345
|
+
│ - Pattern discovery
|
|
150
346
|
│
|
|
151
|
-
|
|
152
|
-
│ "Suggest UI implementation direction"
|
|
347
|
+
↓ (all complete)
|
|
153
348
|
│
|
|
154
|
-
|
|
155
|
-
Opus: Synthesize analysis results, decide implementation direction
|
|
349
|
+
├─→ Opus: Synthesize and decide
|
|
156
350
|
│
|
|
157
|
-
|
|
158
|
-
Task(sonnet): Core implementation
|
|
159
|
-
"Implement code according to SPEC"
|
|
351
|
+
├─→ Task(sonnet): Implementation
|
|
160
352
|
│
|
|
161
|
-
|
|
162
|
-
Task(haiku): Write tests
|
|
163
|
-
"Write tests for implemented code"
|
|
353
|
+
├─→ Task(haiku): Tests
|
|
164
354
|
│
|
|
165
355
|
↓
|
|
166
|
-
|
|
356
|
+
Phase N Complete → Next Phase
|
|
167
357
|
```
|
|
168
358
|
|
|
169
|
-
|
|
170
|
-
```javascript
|
|
171
|
-
// Independent tasks run in parallel
|
|
172
|
-
Task(haiku) - Code analysis
|
|
173
|
-
Task(haiku) - Dependency check
|
|
174
|
-
// → Run simultaneously
|
|
175
|
-
|
|
176
|
-
// Sequential tasks
|
|
177
|
-
Task(sonnet) - Implementation (after analysis complete)
|
|
178
|
-
Task(haiku) - Tests (after implementation complete)
|
|
179
|
-
```
|
|
359
|
+
---
|
|
180
360
|
|
|
181
361
|
1. **Related code analysis**: Task(haiku) explores `<context>` related code
|
|
182
362
|
2. **File creation/modification**: Task(sonnet) implements per `<output_format>`
|
|
@@ -251,12 +431,36 @@ Claude:
|
|
|
251
431
|
|
|
252
432
|
🚀 Starting Phase 1...
|
|
253
433
|
|
|
434
|
+
⚡ Launching parallel exploration...
|
|
435
|
+
[Task(haiku) × 3 launched in parallel]
|
|
436
|
+
- Analyzing Phaser.js project patterns
|
|
437
|
+
- Checking TypeScript configuration standards
|
|
438
|
+
- Finding game loop conventions
|
|
439
|
+
|
|
440
|
+
✅ Exploration complete (3 tasks, 8.2s)
|
|
441
|
+
📋 Synthesizing results...
|
|
442
|
+
|
|
443
|
+
🔨 Implementing Phase 1...
|
|
444
|
+
[Task(sonnet) - Project setup]
|
|
445
|
+
|
|
254
446
|
✅ Phase 1 complete
|
|
255
447
|
- package.json created
|
|
256
448
|
- TypeScript configured
|
|
257
449
|
- Phaser.js installed
|
|
258
450
|
|
|
259
451
|
🚀 Starting Phase 2...
|
|
452
|
+
|
|
453
|
+
⚡ Launching parallel exploration...
|
|
454
|
+
[Task(haiku) × 3 launched in parallel]
|
|
455
|
+
- Analyzing game scene structure
|
|
456
|
+
- Checking physics engine patterns
|
|
457
|
+
- Finding collision detection approaches
|
|
458
|
+
|
|
459
|
+
✅ Exploration complete (3 tasks, 7.5s)
|
|
460
|
+
|
|
461
|
+
🔨 Implementing Phase 2...
|
|
462
|
+
[Task(sonnet) - Game logic]
|
|
463
|
+
|
|
260
464
|
[Implementation continues...]
|
|
261
465
|
|
|
262
466
|
✅ All Phases complete!
|
|
@@ -267,6 +471,8 @@ Claude:
|
|
|
267
471
|
✅ npm run build succeeds
|
|
268
472
|
|
|
269
473
|
🎉 Implementation complete!
|
|
474
|
+
|
|
475
|
+
⏱️ Total time: 4m 32s (vs ~12m sequential)
|
|
270
476
|
```
|
|
271
477
|
|
|
272
478
|
### Phase-specific Execution
|
|
@@ -285,6 +491,9 @@ Phase 2: Game Logic
|
|
|
285
491
|
4. [ ] Score system
|
|
286
492
|
5. [ ] Game over conditions
|
|
287
493
|
|
|
494
|
+
⚡ Launching parallel exploration...
|
|
495
|
+
[Task(haiku) × 3 launched in parallel]
|
|
496
|
+
|
|
288
497
|
🚀 Starting implementation...
|
|
289
498
|
```
|
|
290
499
|
|
package/CLAUDE.md
CHANGED
|
@@ -8,12 +8,29 @@ SPEC 주도 AI 코딩 프레임워크 (Claude Code 전용)
|
|
|
8
8
|
/vibe.spec → /vibe.run → /vibe.verify
|
|
9
9
|
```
|
|
10
10
|
|
|
11
|
+
## ULTRAWORK Mode (권장)
|
|
12
|
+
|
|
13
|
+
`ultrawork` 또는 `ulw` 키워드를 포함하면 최대 성능 모드 활성화:
|
|
14
|
+
|
|
15
|
+
```bash
|
|
16
|
+
/vibe.run "기능명" ultrawork # 모든 최적화 자동 활성화
|
|
17
|
+
/vibe.run "기능명" ulw # 동일 (단축어)
|
|
18
|
+
```
|
|
19
|
+
|
|
20
|
+
**활성화 기능:**
|
|
21
|
+
- 병렬 서브에이전트 탐색 (3+ 동시)
|
|
22
|
+
- Boulder Loop (모든 Phase 완료까지 자동 진행)
|
|
23
|
+
- 에러 자동 재시도 (최대 3회)
|
|
24
|
+
- 컨텍스트 70%+ 시 자동 압축/저장
|
|
25
|
+
- Phase 간 확인 없이 연속 실행
|
|
26
|
+
|
|
11
27
|
## Commands
|
|
12
28
|
|
|
13
29
|
| 명령어 | 설명 |
|
|
14
30
|
|--------|------|
|
|
15
31
|
| `/vibe.spec "기능명"` | SPEC 작성 (PTCF 구조) |
|
|
16
32
|
| `/vibe.run "기능명"` | 구현 실행 |
|
|
33
|
+
| `/vibe.run "기능명" ultrawork` | **최대 성능 모드** |
|
|
17
34
|
| `/vibe.verify "기능명"` | 검증 |
|
|
18
35
|
| `/vibe.reason "문제"` | 체계적 추론 |
|
|
19
36
|
| `/vibe.analyze` | 프로젝트 분석 |
|
package/README.md
CHANGED
|
@@ -13,10 +13,12 @@ SPEC 문서 하나로 AI가 바로 구현하는 2-step 워크플로우.
|
|
|
13
13
|
## Features
|
|
14
14
|
|
|
15
15
|
- **Claude Code 전용**: 네이티브 슬래시 커맨드 + MCP 통합
|
|
16
|
-
- **
|
|
16
|
+
- **ULTRAWORK Mode**: `ulw` 한 단어로 모든 최적화 자동 활성화
|
|
17
|
+
- **Boulder Loop**: 모든 Phase 완료까지 자동 진행 (시지푸스처럼)
|
|
18
|
+
- **병렬 서브에이전트**: Haiku 3+ 동시 탐색으로 ~3배 속도 향상
|
|
17
19
|
- **PTCF 구조**: Persona, Task, Context, Format - Gemini 프롬프트 최적화
|
|
18
20
|
- **코딩 규칙 내장**: `.vibe/rules/` - 품질 기준, 복잡도 제한, 안티패턴
|
|
19
|
-
- **자동
|
|
21
|
+
- **자동 컨텍스트 관리**: 70%+ 시 자동 압축/저장
|
|
20
22
|
- **36개 MCP 도구**: @su-record/hi-ai 통합
|
|
21
23
|
|
|
22
24
|
---
|
|
@@ -46,9 +48,10 @@ cd my-project
|
|
|
46
48
|
### 2. 슬래시 커맨드 사용 (Claude Code)
|
|
47
49
|
|
|
48
50
|
```
|
|
49
|
-
/vibe.spec "로그인 기능"
|
|
50
|
-
/vibe.run "로그인 기능"
|
|
51
|
-
/vibe.
|
|
51
|
+
/vibe.spec "로그인 기능" # SPEC 작성 (대화형)
|
|
52
|
+
/vibe.run "로그인 기능" # 구현 실행
|
|
53
|
+
/vibe.run "로그인 기능" ultrawork # 🚀 최대 성능 모드 (권장)
|
|
54
|
+
/vibe.verify "로그인 기능" # 검증
|
|
52
55
|
```
|
|
53
56
|
|
|
54
57
|
---
|
|
@@ -62,10 +65,10 @@ cd my-project
|
|
|
62
65
|
│ ↓ .vibe/specs/{기능명}.md (PTCF 구조) │
|
|
63
66
|
│ ↓ .vibe/features/{기능명}.feature (BDD) │
|
|
64
67
|
├─────────────────────────────────────────────────────┤
|
|
65
|
-
│ /vibe.run "기능명"
|
|
66
|
-
│ ↓
|
|
67
|
-
│ ↓ Phase
|
|
68
|
-
│ ↓
|
|
68
|
+
│ /vibe.run "기능명" ultrawork │
|
|
69
|
+
│ ↓ 🚀 ULTRAWORK: 병렬 탐색 + 자동 진행 │
|
|
70
|
+
│ ↓ Boulder Loop: 모든 Phase 완료까지 자동 실행 │
|
|
71
|
+
│ ↓ 에러 자동 재시도 (3회) │
|
|
69
72
|
├─────────────────────────────────────────────────────┤
|
|
70
73
|
│ /vibe.verify "기능명" │
|
|
71
74
|
│ ↓ Acceptance Criteria 검증 │
|
|
@@ -105,6 +108,8 @@ cd my-project
|
|
|
105
108
|
|--------|------|
|
|
106
109
|
| `/vibe.spec "기능명"` | SPEC 작성 (PTCF 구조) |
|
|
107
110
|
| `/vibe.run "기능명"` | 구현 실행 |
|
|
111
|
+
| `/vibe.run "기능명" ultrawork` | 🚀 **최대 성능 모드** (권장) |
|
|
112
|
+
| `/vibe.run "기능명" ulw` | ultrawork 단축어 |
|
|
108
113
|
| `/vibe.run "기능명" --phase N` | 특정 Phase만 실행 |
|
|
109
114
|
| `/vibe.verify "기능명"` | 검증 |
|
|
110
115
|
|
|
@@ -125,6 +130,52 @@ cd my-project
|
|
|
125
130
|
|
|
126
131
|
---
|
|
127
132
|
|
|
133
|
+
## ULTRAWORK Mode
|
|
134
|
+
|
|
135
|
+
> `ultrawork` 또는 `ulw` 키워드 하나로 **모든 최적화가 자동 활성화**됩니다.
|
|
136
|
+
|
|
137
|
+
```bash
|
|
138
|
+
/vibe.run "기능명" ultrawork # 최대 성능 모드
|
|
139
|
+
/vibe.run "기능명" ulw # 동일 (단축어)
|
|
140
|
+
```
|
|
141
|
+
|
|
142
|
+
### 활성화 기능
|
|
143
|
+
|
|
144
|
+
| 기능 | 설명 |
|
|
145
|
+
|------|------|
|
|
146
|
+
| **병렬 탐색** | 3+ Task(haiku) 에이전트 동시 실행 |
|
|
147
|
+
| **Boulder Loop** | 모든 Phase 완료까지 자동 진행 (멈추지 않음) |
|
|
148
|
+
| **자동 재시도** | 에러 발생 시 최대 3회 자동 재시도 |
|
|
149
|
+
| **컨텍스트 관리** | 70%+ 시 자동 압축 및 저장 |
|
|
150
|
+
| **무중단 실행** | Phase 간 확인 없이 연속 진행 |
|
|
151
|
+
| **외부 LLM** | GPT/Gemini 활성화 시 자동 참조 |
|
|
152
|
+
|
|
153
|
+
### Boulder Loop
|
|
154
|
+
|
|
155
|
+
시지푸스처럼 바위를 굴리듯, **모든 작업이 완료될 때까지** 자동으로 진행:
|
|
156
|
+
|
|
157
|
+
```
|
|
158
|
+
Phase 1 → Phase 2 → Phase 3 → ... → Phase N
|
|
159
|
+
↓ ↓ ↓ ↓
|
|
160
|
+
[병렬탐색] [병렬탐색] [병렬탐색] [병렬탐색]
|
|
161
|
+
[구현] [구현] [구현] [구현]
|
|
162
|
+
[테스트] [테스트] [테스트] [테스트]
|
|
163
|
+
↓
|
|
164
|
+
🎉 완료까지 자동!
|
|
165
|
+
```
|
|
166
|
+
|
|
167
|
+
### 일반 모드 vs ULTRAWORK
|
|
168
|
+
|
|
169
|
+
| 항목 | 일반 모드 | ULTRAWORK |
|
|
170
|
+
|------|----------|-----------|
|
|
171
|
+
| Phase 전환 | 일시정지 가능 | 자동 진행 |
|
|
172
|
+
| 에러 발생 | 보고 후 중단 | 자동 재시도 (3회) |
|
|
173
|
+
| 컨텍스트 70%+ | 경고만 | 자동 압축/저장 |
|
|
174
|
+
| 탐색 방식 | 순차 가능 | **강제 병렬** |
|
|
175
|
+
| 완료 조건 | Phase별 | 전체 완료까지 |
|
|
176
|
+
|
|
177
|
+
---
|
|
178
|
+
|
|
128
179
|
## Project Structure
|
|
129
180
|
|
|
130
181
|
`vibe init` 실행 후 생성되는 구조:
|
|
@@ -251,8 +302,11 @@ AI의 역할과 전문성 정의
|
|
|
251
302
|
| Hook | 트리거 | 동작 |
|
|
252
303
|
|------|--------|------|
|
|
253
304
|
| `SessionStart` | 세션 시작 | 이전 컨텍스트 자동 복원 |
|
|
254
|
-
| `
|
|
255
|
-
| `
|
|
305
|
+
| `UserPromptSubmit` | `ultrawork`/`ulw` 감지 | 🚀 ULTRAWORK 모드 활성화 |
|
|
306
|
+
| `PostToolUse` | Write/Edit 후 | 품질 체크 (멈추지 않음) |
|
|
307
|
+
| `Notification` | 컨텍스트 70% | 경고 (ULTRAWORK: 자동 저장) |
|
|
308
|
+
| `Notification` | 컨텍스트 85% | 즉시 저장 + 압축 |
|
|
309
|
+
| `Notification` | 컨텍스트 95% | 긴급 저장 + 세션 전환 준비 |
|
|
256
310
|
|
|
257
311
|
### 선택적 연동 (외부 LLM)
|
|
258
312
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@su-record/vibe",
|
|
3
|
-
"version": "1.1
|
|
3
|
+
"version": "1.2.1",
|
|
4
4
|
"description": "Vibe - Claude Code exclusive SPEC-driven AI coding framework",
|
|
5
5
|
"bin": {
|
|
6
6
|
"vibe": "./bin/vibe"
|
|
@@ -18,7 +18,9 @@
|
|
|
18
18
|
"mcp",
|
|
19
19
|
"framework",
|
|
20
20
|
"requirements",
|
|
21
|
-
"user-story"
|
|
21
|
+
"user-story",
|
|
22
|
+
"ultrawork",
|
|
23
|
+
"parallel-agents"
|
|
22
24
|
],
|
|
23
25
|
"author": "Su",
|
|
24
26
|
"license": "MIT",
|
|
@@ -10,13 +10,24 @@
|
|
|
10
10
|
]
|
|
11
11
|
}
|
|
12
12
|
],
|
|
13
|
+
"UserPromptSubmit": [
|
|
14
|
+
{
|
|
15
|
+
"matcher": "ultrawork|ulw|울트라워크",
|
|
16
|
+
"hooks": [
|
|
17
|
+
{
|
|
18
|
+
"type": "prompt",
|
|
19
|
+
"prompt": "🚀 ULTRAWORK MODE DETECTED. You MUST: (1) Use PARALLEL Task calls for ALL exploration, (2) Enable Boulder Loop - auto-continue through ALL phases without pausing, (3) Auto-retry on errors up to 3 times, (4) Run mcp__vibe__auto_save_context at 70%+ context, (5) Do NOT ask for confirmation between phases. Keep rolling until ALL acceptance criteria pass."
|
|
20
|
+
}
|
|
21
|
+
]
|
|
22
|
+
}
|
|
23
|
+
],
|
|
13
24
|
"PostToolUse": [
|
|
14
25
|
{
|
|
15
26
|
"matcher": "Write|Edit",
|
|
16
27
|
"hooks": [
|
|
17
28
|
{
|
|
18
29
|
"type": "prompt",
|
|
19
|
-
"prompt": "
|
|
30
|
+
"prompt": "Silently check .vibe/rules/quality/checklist.md. Do NOT stop or report - continue your current task. Only fix critical issues (syntax errors, security vulnerabilities) inline. Minor issues (lint warnings, style) can be addressed at the end of the current phase."
|
|
20
31
|
}
|
|
21
32
|
]
|
|
22
33
|
}
|
|
@@ -27,7 +38,7 @@
|
|
|
27
38
|
"hooks": [
|
|
28
39
|
{
|
|
29
40
|
"type": "prompt",
|
|
30
|
-
"prompt": "Context at 70%.
|
|
41
|
+
"prompt": "Context at 70%. You still have headroom - don't rush or produce sloppy work. But start being selective about file reads. If ULTRAWORK mode: run mcp__vibe__auto_save_context now."
|
|
31
42
|
}
|
|
32
43
|
]
|
|
33
44
|
},
|
|
@@ -36,7 +47,7 @@
|
|
|
36
47
|
"hooks": [
|
|
37
48
|
{
|
|
38
49
|
"type": "prompt",
|
|
39
|
-
"prompt": "Context at 85%.
|
|
50
|
+
"prompt": "Context at 85%. IMMEDIATELY run mcp__vibe__auto_save_context with urgency='high'. Summarize current progress and next steps. Compress verbose outputs."
|
|
40
51
|
}
|
|
41
52
|
]
|
|
42
53
|
},
|
|
@@ -45,7 +56,7 @@
|
|
|
45
56
|
"hooks": [
|
|
46
57
|
{
|
|
47
58
|
"type": "prompt",
|
|
48
|
-
"prompt": "Context at 95
|
|
59
|
+
"prompt": "Context at 95% - CRITICAL. Run mcp__vibe__auto_save_context with urgency='critical' NOW. Save all progress, blockers, and next steps. Prepare for session transition. If in ULTRAWORK mode, the next session will auto-restore and continue."
|
|
49
60
|
}
|
|
50
61
|
]
|
|
51
62
|
}
|