oh-my-customcode 0.76.0 → 0.76.2
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 +3 -3
- package/dist/cli/index.js +1 -1
- package/dist/index.js +1 -1
- package/package.json +1 -1
- package/templates/.claude/skills/agora/SKILL.md +194 -0
- package/templates/CLAUDE.md +2 -1
- package/templates/manifest.json +2 -2
package/README.md
CHANGED
|
@@ -13,7 +13,7 @@
|
|
|
13
13
|
|
|
14
14
|
**[한국어 문서 (Korean)](./README_ko.md)**
|
|
15
15
|
|
|
16
|
-
46 agents.
|
|
16
|
+
46 agents. 101 skills. 21 rules. One command.
|
|
17
17
|
|
|
18
18
|
```bash
|
|
19
19
|
npm install -g oh-my-customcode && cd your-project && omcustom init
|
|
@@ -145,7 +145,7 @@ Each agent declares its tools, model, memory scope, and limitations in YAML fron
|
|
|
145
145
|
|
|
146
146
|
---
|
|
147
147
|
|
|
148
|
-
### Skills (
|
|
148
|
+
### Skills (101)
|
|
149
149
|
|
|
150
150
|
| Category | Count | Includes |
|
|
151
151
|
|----------|-------|----------|
|
|
@@ -285,7 +285,7 @@ your-project/
|
|
|
285
285
|
├── CLAUDE.md # Entry point
|
|
286
286
|
├── .claude/
|
|
287
287
|
│ ├── agents/ # 46 agent definitions
|
|
288
|
-
│ ├── skills/ #
|
|
288
|
+
│ ├── skills/ # 101 skill modules
|
|
289
289
|
│ ├── rules/ # 21 governance rules (R000-R021)
|
|
290
290
|
│ ├── hooks/ # 15 lifecycle hook scripts
|
|
291
291
|
│ ├── schemas/ # Tool input validation schemas
|
package/dist/cli/index.js
CHANGED
package/dist/index.js
CHANGED
package/package.json
CHANGED
|
@@ -0,0 +1,194 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: omcustom:agora
|
|
3
|
+
description: "Multi-LLM adversarial consensus loop — 3+ LLMs compete to find flaws in designs/specs until unanimous agreement is reached"
|
|
4
|
+
user-invocable: true
|
|
5
|
+
argument-hint: "<document-path> [--rounds N] [--severity-threshold HIGH]"
|
|
6
|
+
effort: max
|
|
7
|
+
scope: core
|
|
8
|
+
version: 1.0.0
|
|
9
|
+
source:
|
|
10
|
+
type: external
|
|
11
|
+
origin: github
|
|
12
|
+
url: https://github.com/baekenough/baekenough-skills
|
|
13
|
+
version: 1.0.0
|
|
14
|
+
---
|
|
15
|
+
|
|
16
|
+
# Agora: Multi-LLM Adversarial Consensus
|
|
17
|
+
|
|
18
|
+
3개 이상의 LLM(Claude, Codex/GPT, Gemini)이 경쟁적으로 설계/문서의 결함을 찾고, 만장일치 합의에 도달할 때까지 반복하는 적대적 교차 검증 스킬.
|
|
19
|
+
|
|
20
|
+
## Prerequisites
|
|
21
|
+
|
|
22
|
+
- `codex-exec` skill (Codex/GPT 호출)
|
|
23
|
+
- `gemini-exec` skill (Gemini 호출)
|
|
24
|
+
- Agent Teams enabled (`CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`) or Agent tool available
|
|
25
|
+
|
|
26
|
+
## Usage
|
|
27
|
+
|
|
28
|
+
```
|
|
29
|
+
/agora docs/design.md # Default: 3 LLMs, unlimited rounds
|
|
30
|
+
/agora docs/design.md --rounds 10 # Max 10 rounds
|
|
31
|
+
/agora docs/design.md --severity-threshold HIGH # Exit when no HIGH+ findings
|
|
32
|
+
/agora docs/design.md --models claude,codex # 2 LLMs only
|
|
33
|
+
```
|
|
34
|
+
|
|
35
|
+
## Workflow
|
|
36
|
+
|
|
37
|
+
### Phase 1: Setup
|
|
38
|
+
1. Read the target document
|
|
39
|
+
2. Create Agent Team: `TeamCreate("agora-review")`
|
|
40
|
+
3. Create review tasks per focus area
|
|
41
|
+
|
|
42
|
+
### Phase 2: Spawn Reviewers (parallel)
|
|
43
|
+
Spawn 3 reviewers as Agent Team members:
|
|
44
|
+
|
|
45
|
+
```
|
|
46
|
+
Agent(name: "claude-critic", model: opus, effort: max)
|
|
47
|
+
→ 20-point deep adversarial review
|
|
48
|
+
|
|
49
|
+
Agent(name: "codex-critic", model: opus)
|
|
50
|
+
→ Invoke Skill(codex-exec) for GPT perspective + independent Claude analysis
|
|
51
|
+
|
|
52
|
+
Agent(name: "gemini-critic", model: opus)
|
|
53
|
+
→ Invoke Skill(gemini-exec) for Gemini perspective + independent Claude analysis
|
|
54
|
+
```
|
|
55
|
+
|
|
56
|
+
### Phase 3: Independent Review
|
|
57
|
+
Each reviewer performs adversarial review with this template:
|
|
58
|
+
|
|
59
|
+
```
|
|
60
|
+
For EACH review point:
|
|
61
|
+
### Round N: [Topic]
|
|
62
|
+
**Severity**: CRITICAL / HIGH / MEDIUM / LOW
|
|
63
|
+
**Flaw**: [Specific, concrete problem description]
|
|
64
|
+
**Evidence**: [Why this is real, not theoretical]
|
|
65
|
+
**Impact**: [What happens if not addressed]
|
|
66
|
+
**Counter-argument**: [Best case FOR the current design]
|
|
67
|
+
**Verdict**: KEEP / MODIFY / REJECT
|
|
68
|
+
```
|
|
69
|
+
|
|
70
|
+
Review areas (adapt to document type):
|
|
71
|
+
- Architecture fundamentals
|
|
72
|
+
- Component/service design
|
|
73
|
+
- Data architecture
|
|
74
|
+
- Security & resilience
|
|
75
|
+
- Feasibility & deployment
|
|
76
|
+
- Testing strategy
|
|
77
|
+
- Operational complexity
|
|
78
|
+
|
|
79
|
+
### Phase 4: Cross-Review (Peer-to-Peer)
|
|
80
|
+
Each reviewer sends findings to the other two via `SendMessage`.
|
|
81
|
+
|
|
82
|
+
Counter-review template:
|
|
83
|
+
1. Which findings do you **AGREE** with? (and why)
|
|
84
|
+
2. Which findings do you **DISAGREE** with? (evidence-based rebuttal)
|
|
85
|
+
3. What did they **MISS** that you caught?
|
|
86
|
+
4. What did they catch that you **MISSED**?
|
|
87
|
+
5. **SEVERITY** adjustments — upgrade or downgrade with justification
|
|
88
|
+
|
|
89
|
+
### Phase 5: Synthesis
|
|
90
|
+
Team lead aggregates all findings:
|
|
91
|
+
|
|
92
|
+
```
|
|
93
|
+
UNANIMOUS CRITICAL: [findings all 3 agreed on]
|
|
94
|
+
STRONG AGREEMENT: [findings 2/3 agreed on]
|
|
95
|
+
SPLIT DECISIONS: [findings with disagreement + resolution]
|
|
96
|
+
```
|
|
97
|
+
|
|
98
|
+
Determine verdict:
|
|
99
|
+
- **BUILD**: No CRITICAL, no unresolved HIGH
|
|
100
|
+
- **BUILD WITH CHANGES**: No CRITICAL, HIGH findings have accepted mitigations
|
|
101
|
+
- **REDESIGN**: Any unresolved CRITICAL findings
|
|
102
|
+
- **ABANDON**: Fundamental concept is flawed
|
|
103
|
+
|
|
104
|
+
### Phase 6: Loop (if REDESIGN)
|
|
105
|
+
1. Team lead produces/delegates redesign addressing ALL critical findings
|
|
106
|
+
2. New version sent to ALL reviewers: `SendMessage(to: "*")`
|
|
107
|
+
3. Reviewers re-review → GOTO Phase 4
|
|
108
|
+
4. Repeat until EXIT criteria met
|
|
109
|
+
|
|
110
|
+
### Phase 7: Exit (consensus reached)
|
|
111
|
+
When ALL reviewers agree BUILD or BUILD WITH CHANGES:
|
|
112
|
+
1. Produce final consensus report
|
|
113
|
+
2. Write to `.claude/outputs/sessions/{date}/agora-{topic}-{time}.md`
|
|
114
|
+
3. Shut down team: `SendMessage(to: "*", message: {type: "shutdown_request"})`
|
|
115
|
+
|
|
116
|
+
## Reviewer Principles
|
|
117
|
+
|
|
118
|
+
1. **NEUTRAL** — no reviewer has home team advantage
|
|
119
|
+
2. **COMPETITIVE** — find flaws others missed
|
|
120
|
+
3. **CRITICAL** — "fewer than 5 CRITICAL flaws = not looking hard enough"
|
|
121
|
+
4. **EVIDENCE-BASED** — every finding cites specific evidence
|
|
122
|
+
5. **CONSTRUCTIVE** — every flaw includes recommended fix
|
|
123
|
+
6. **CONVERGENT** — goal is consensus, not endless disagreement
|
|
124
|
+
|
|
125
|
+
## Consensus Criteria
|
|
126
|
+
|
|
127
|
+
| Condition | Required |
|
|
128
|
+
|-----------|----------|
|
|
129
|
+
| CRITICAL findings resolved | ALL |
|
|
130
|
+
| HIGH findings resolved or accepted | ALL |
|
|
131
|
+
| All reviewers rate BUILD or BUILD WITH CHANGES | YES |
|
|
132
|
+
| Cross-review disagreements resolved | ALL |
|
|
133
|
+
|
|
134
|
+
## Output Format
|
|
135
|
+
|
|
136
|
+
```markdown
|
|
137
|
+
# Agora Consensus Report
|
|
138
|
+
|
|
139
|
+
## Document: [path]
|
|
140
|
+
## Rounds: [N]
|
|
141
|
+
## Reviewers: [list with LLM models used]
|
|
142
|
+
|
|
143
|
+
## Verdict: [BUILD / BUILD WITH CHANGES / REDESIGN]
|
|
144
|
+
|
|
145
|
+
## Unanimous Findings
|
|
146
|
+
| # | Finding | Severity | All 3 Agree |
|
|
147
|
+
|---|---------|----------|-------------|
|
|
148
|
+
|
|
149
|
+
## Required Changes Before Build
|
|
150
|
+
1. [change with source reviewer]
|
|
151
|
+
2. ...
|
|
152
|
+
|
|
153
|
+
## Accepted Risks
|
|
154
|
+
- [finding accepted with justification]
|
|
155
|
+
|
|
156
|
+
## Unique Contributions Per Reviewer
|
|
157
|
+
| Reviewer | Findings Others Missed |
|
|
158
|
+
|----------|----------------------|
|
|
159
|
+
|
|
160
|
+
## Process Metrics
|
|
161
|
+
- Rounds: N
|
|
162
|
+
- Total findings: N
|
|
163
|
+
- Cross-adopted: N
|
|
164
|
+
- Severity upgrades: N
|
|
165
|
+
- Severity downgrades: N
|
|
166
|
+
- Disagreements raised: N
|
|
167
|
+
- Disagreements resolved: N/N
|
|
168
|
+
```
|
|
169
|
+
|
|
170
|
+
## Configuration
|
|
171
|
+
|
|
172
|
+
```yaml
|
|
173
|
+
# Default settings
|
|
174
|
+
agora:
|
|
175
|
+
max_rounds: unlimited # Set --rounds to limit
|
|
176
|
+
severity_threshold: HIGH # EXIT when no findings >= threshold
|
|
177
|
+
models:
|
|
178
|
+
- claude (opus, max effort)
|
|
179
|
+
- codex (via codex-exec skill)
|
|
180
|
+
- gemini (via gemini-exec skill)
|
|
181
|
+
review_points: 20 # Per reviewer
|
|
182
|
+
cross_review: true # Peer-to-peer sharing
|
|
183
|
+
auto_redesign: true # Auto-produce redesign on REDESIGN verdict
|
|
184
|
+
```
|
|
185
|
+
|
|
186
|
+
## Anti-Patterns
|
|
187
|
+
|
|
188
|
+
| Anti-Pattern | Why Wrong | Correct |
|
|
189
|
+
|-------------|-----------|---------|
|
|
190
|
+
| Single LLM review | Misses blind spots | 3+ LLMs find complementary flaws |
|
|
191
|
+
| No cross-review | Reviewers don't challenge each other | Peer-to-peer sharing surfaces disagreements |
|
|
192
|
+
| Accepting first BUILD | May miss edge cases | Loop until ALL agree |
|
|
193
|
+
| Ignoring split decisions | Unresolved disagreements fester | Resolve every split with evidence |
|
|
194
|
+
| Push for consensus too fast | Premature agreement | Let reviewers challenge freely |
|
package/templates/CLAUDE.md
CHANGED
|
@@ -108,6 +108,7 @@ oh-my-customcode로 구동됩니다.
|
|
|
108
108
|
| `/omcustom:improve-report` | eval-core 기반 개선 현황 리포트 |
|
|
109
109
|
| `/omcustom-takeover` | 기존 에이전트/스킬에서 canonical spec 추출 |
|
|
110
110
|
| `/adversarial-review` | 공격자 관점 보안 코드 리뷰 |
|
|
111
|
+
| `/omcustom:agora` | Multi-LLM 적대적 합의 루프 (설계/스펙 교차 검증) |
|
|
111
112
|
| `/ambiguity-gate` | 요청 모호성 분석 및 명확화 질문 (ouroboros 패턴) |
|
|
112
113
|
| `/dev-review` | 코드 베스트 프랙티스 리뷰 |
|
|
113
114
|
| `/dev-refactor` | 코드 리팩토링 |
|
|
@@ -151,7 +152,7 @@ project/
|
|
|
151
152
|
+-- CLAUDE.md # 진입점
|
|
152
153
|
+-- .claude/
|
|
153
154
|
| +-- agents/ # 서브에이전트 정의 (46 파일)
|
|
154
|
-
| +-- skills/ # 스킬 (
|
|
155
|
+
| +-- skills/ # 스킬 (100 디렉토리)
|
|
155
156
|
| +-- rules/ # 전역 규칙 (R000-R021)
|
|
156
157
|
| +-- hooks/ # 훅 스크립트 (보안, 검증, HUD)
|
|
157
158
|
| +-- contexts/ # 컨텍스트 파일 (ecomode)
|
package/templates/manifest.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
{
|
|
2
|
-
"version": "0.76.
|
|
2
|
+
"version": "0.76.2",
|
|
3
3
|
"lastUpdated": "2026-03-24T00:00:00.000Z",
|
|
4
4
|
"components": [
|
|
5
5
|
{
|
|
@@ -18,7 +18,7 @@
|
|
|
18
18
|
"name": "skills",
|
|
19
19
|
"path": ".claude/skills",
|
|
20
20
|
"description": "Reusable skill modules (includes slash commands)",
|
|
21
|
-
"files":
|
|
21
|
+
"files": 101
|
|
22
22
|
},
|
|
23
23
|
{
|
|
24
24
|
"name": "guides",
|