跳转至

原文快照

这里按编号列出 Claude 官方资料的原文快照。每条链接指向本地文件，原始 URL 已写在文件头部。编号与 references.md 对照；第 10 章的论文与实验只在 references.md 提供外部链接，不做快照。

  1. Agent Skills
  2. Get started with Agent Skills in the API
  3. Skill authoring best practices
  4. Tool use with Claude
  5. How to implement tool use
  6. Programmatic tool calling
  7. Fine-grained tool streaming
  8. Tool search tool
  9. Web search tool
  10. Web fetch tool
  11. Memory tool
  12. Code execution tool
  13. Bash tool
  14. Text editor tool
  15. Computer use tool
  16. MCP connector
  17. Remote MCP servers
  18. What is the Model Context Protocol (MCP)? - Model Context Protocol
  19. Agent SDK overview
  20. Quickstart
  21. Agent SDK reference - Python
  22. Agent SDK reference - TypeScript
  23. TypeScript SDK V2 interface (preview)
  24. Agent Skills in the SDK
  25. Subagents in the SDK
  26. Intercept and control agent behavior with hooks
  27. Configure permissions
  28. Custom Tools
  29. MCP in the SDK
  30. Structured outputs in the SDK
  31. Tracking Costs and Usage
  32. Rewind file changes with checkpointing
  33. Streaming Input
  34. Session Management
  35. Slash Commands in the SDK
  36. Plugins in the SDK
  37. Modifying system prompts
  38. Hosting the Agent SDK
  39. Securely deploying AI agents
  40. Migrate to Claude Agent SDK
  41. Todo Lists
  42. Building Effective AI Agents
  43. Our framework for developing safe and trustworthy agents
  44. Introducing the Model Context Protocol
  45. Donating the Model Context Protocol and establishing the Agentic AI Foundation
  46. Developing a computer use model
  47. Introducing computer use, a new Claude 3.5 Sonnet, and Claude 3.5 Haiku
  48. Enabling Claude Code to work more autonomously
  49. Claude Code and new admin controls for business plans
  50. Anthropic acquires Bun as Claude Code reaches $1B milestone
  51. Agentic Misalignment: How LLMs could be insider threats
  52. Simple probes can catch sleeper agents
  53. Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
  54. Introducing advanced tool use on the Claude Developer Platform
  55. Building agents with the Claude Agent SDK
  56. Claude Code Best Practices
  57. Making Claude Code more secure and autonomous with sandboxing
  58. The "think" tool: Enabling Claude to stop and think
  59. Code execution with MCP: building more efficient AI agents
  60. Claude Desktop Extensions: One-click MCP server installation for Claude Desktop
  61. Effective context engineering for AI agents
  62. Effective harnesses for long-running agents
  63. Equipping agents for the real world with Agent Skills
  64. How we built our multi-agent research system
  65. Writing effective tools for AI agents - using AI agents
  66. Demystifying evals for AI agents