● 社群
nyldn
🔑 需要你的金鑰
Claude Octopus
作者 nyldn · nyldn/claude-octopus
Run up to 8 models in parallel on the same task; surface where they disagree — that's where your blindspot is.
Octopus is a multi-model orchestration skill: define a task, it dispatches to N model providers, then highlights consensus and disagreement. Useful for research, design review, and plan critique where one model can have a confident-but-wrong opinion.
為什麼要用
核心特性
- Up to 8 models in parallel
- Disagreement highlighting
- Cost guard (caps spend per run)
- Supports Anthropic, OpenAI, Google, xAI, etc.
- Output is a single decision-aid Markdown
使用場景
實戰用法： Claude Octopus
何時使用： You're about to commit to an architecture and want a second (and third, and fourth) opinion.
步驟
-
Brief
Provide the design doc + question.✓ 已複製
→ Octopus dispatches to N models
-
Read disagreements
Skim only where models disagree.✓ 已複製
→ Real risks surface
-
Decide
Resolve disagreements with a focused follow-up.✓ 已複製
→ Documented decision
結果： Decision made with explicit awareness of where models disagree.
組合
與其他 MCP 搭配，撬動十倍槓桿
Use Octopus on the proposed plan, then flow-next executes the agreed-on version
Combine claude-octopus with flow-next-skill: Use Octopus on the proposed plan, then flow-next executes the agreed-on version✓ 已複製
成本與限制
運行它的成本
- API 配額
- Per-provider limits
- 每次呼叫 Token 數
- N × normal
- 費用
- Pay-per-call across providers
- 提示
- Set the cost cap; use cheap models for the panel and one premium
安全
權限、密鑰、影響範圍
最小權限： multi-provider keys
憑證儲存： API keys per provider in env
資料出站： Each configured provider
切勿授予： ship customer data through every provider blindly — review egress
故障排查
常見錯誤與修復
One provider 401s
Octopus continues with the rest; log shows which key
替代方案
Claude Octopus 對比其他方案
| 替代方案 | 何時用它替代 | 權衡 |
|---|
| Single-model with self-critique | Cost-sensitive | Misses provider-specific blindspots |