🤖 Claude & AI Engineering: Theory to Agents

NOTE

Module Status: Actively expanding. The Anthropic Claude family (Claude 3.5 Sonnet, Claude 3 Opus, Claude 3.7) has set the standard in software reasoning, instruction adherence, and extended context window execution (200k+ tokens).

🧭 Claude Interaction Architecture

graph TD
    Prompt["📝 System Prompt & Context"] --> Claude["🧠 Claude LLM Engine (Anthropic)"]
    Claude --> Artifacts["📄 Artifacts (Code, Markdown, SVGs, HTML)"]
    Claude --> ToolCalls["🛠️ Tool Use / MCP (Model Context Protocol)"]
    Claude --> Subagents["👥 Subagent Orchestration"]
    ToolCalls --> Env["💻 Local OS / Terminal / Web APIs"]

🗺️ Topics & Module Roadmap

1. Claude Model Family & Constitutional AI:
   • Model selection trade-offs (Sonnet for code, Opus for complex reasoning, Haiku for speed/cost).
2. Advanced Prompt Engineering for Codebases:
   • XML tag delimiters (<context>, <guidelines>, <instructions>).
   • Chain-of-Thought prompting, deterministic verification, and hallucination reduction.
3. Context Window Optimization & Prompt Caching:
   • Up to 90% cost reduction and 85% latency reduction using Anthropic Prompt Caching.
4. Claude Code CLI & Pair Programming:
   • Terminal-native agent workflows for multi-file refactoring and test generation.
5. Tool Use & MCP (Model Context Protocol):
   • Connecting Claude models to external servers, filesystems, and databases.

📚 Official Documentation & References

• 🌐 Anthropic Claude Documentation — Official Anthropic guides & Prompt Engineering.
• 🤖 Model Context Protocol (MCP) — Official protocol standard and SDKs.
• 💻 Claude Code CLI Documentation — Claude Code manual.

🔗 Second Brain Links

• Autonomous agent skill authoring in Hermes Agent & Skills.
• Integrate AI in GitHub workflows with Workflows: Issues, PRs & Governance.
• Back to the central hub in Second Brain Docs Hub.