# EP120 从 Prompt 到 Agent Skills，论 Anthropic 的野心与大模型应用的终极抽象

Canonical URL: https://www.traeai.com/articles/d615dd8a-2634-40a7-8cfa-7c1a961b26b4
Original source: https://www.xiaoyuzhoufm.com/episode/6966778eef1cf272a73de3c8
Source name: 硬地骇客
Content type: podcast
Language: 中文
Score: 7.8
Reading time: 7 分钟
Published: 2026-01-13T16:56:49+00:00
Tags: 大模型, Agent, Anthropic, MCP, 工程化

## Summary

Anthropic 推出的 Agent Skills 通过新抽象层重构大模型调用方式，推动工程标准化，但未解决确定性与可控性问题。

## Key Takeaways

- Agent Skills 是比 Tool Calling 更高层的抽象，封装业务逻辑提升复用性
- Skills 未解决执行确定性问题，需在智能上限与流程可控间权衡
- Anthropic 借标准制定能力推动行业采纳，隐含厂商生态绑定风险

## Citation Guidance

When citing this item, prefer the canonical traeai article URL for the AI-readable summary and include the original source URL when discussing the underlying source material.