Tag: Claude Agent SDK
Posts with this tag
- 長任務代理的 Harness：跨上下文穩定交付
根據 Anthropic《Effective harnesses for long-running agents》整理：用初始化代理（initializer）＋漸進式編碼代理（coding）＋特徵清單與端到端測試，讓代理能在多個 context window 間持續推進並保持乾淨狀態。
Looking for an AI platform or Agentic AI partner? I help teams ship enterprise-grade RAG, multi-agent, and real-time AI systems.
Contact