Claude Code

What it is

Anthropic’s official agentic CLI. The most production-hardened agent in the corpus, and the one to read first if you want to see what “shipping a tool that millions use” looks like.

What’s worth studying

Claude Code is the cleanest reference for a generator-driven agent loop with everything turned on: streaming tool calls, explicit prompt caching with a sentinel-string boundary, thinking blocks with signature preservation across turns, latched sticky flags for cache coherence, and a slash-command + skills system that lets users (and Anthropic) extend the agent without touching its core.

Three patterns are uniquely visible here:

1. The cache-edit boundary as a literal string. A 24-character sentinel splits the system prompt into org-static and session-dynamic regions. Survives every refactor and every contributor. See cache-boundary-sentinel.
2. Latched sticky flags. User-toggleable flags that live in the cacheable prefix are latched on: once on, they stay on for the session, even if the user toggles them off. The team accepted a small UX surprise for a 50–70K-token cache win. See latched-sticky-flags.
3. Thinking-signature preservation across turns, stripped on model switch. Anthropic’s HMAC signatures are model-bound; switching from Sonnet to Opus mid-session would crash if Claude Code didn’t strip the blocks. See thinking-signature-preservation.

Drill-down

The full per-doc analysis lives below — these are the original numbered analyses, rendered as styled HTML. Pick a section to study deeper.

• Claude Code — Source Tree Analysis
• 1. Purpose, Problem, and Tech Stack
• 2. Architecture Overview
• 3. Entry Points and Execution Flow
• 4. The Agent Loop and LLM Usage
• 5. The Tool System
• 6. Agents, Skills, Slash Commands, MCP
• 7. Prompts and Guardrails (verbatim)
• 8. Memory and Context
• 9. State, UI, and Session Persistence
• 10. Design Patterns, Tradeoffs, and Clever Tricks
• 11. Key Files Map
• 12. Deep Dive — The /btw Command
• 13. Multi-Agent Coordination
• 14. Mid-Turn Messages, Interruption, and the Command Queue
• 15. Memory System — Deep Dive
• 16. The Agentic Loop — Schematic Walkthrough