中文 | English
Yes. Start with examples/ for intuition, then read docs/layers/, and use docs/source-map.en.md plus docs/source-navigation.en.md to connect that intuition to source. The goal is not to master syntax first, but to understand what each subsystem is solving.
You can start with l1, l3, l4, l5, l6, l7, and l9. For l2 and l8, read the matching layer docs and the source map first, then search the suggested symbols in the source.
Because Claude Code was not designed to be learned linearly by directory order. A better way is to read by problem, call chain, and design decision.
If you are new, read query.ts first to understand the loop. Then move to QueryEngine.ts to understand how system prompts, skills, plugins, and slash commands are staged before the loop begins. The most reliable path is still: example -> layer -> source map -> source navigation.
The first 9 layers are the main course with runnable examples. L10-L15 are source-heavy deep-dive topics for readers who are already inside the codebase.
Start with the Learning Paths and move through P1 -> P2 -> P3 -> P4. At each stage, run the matching example, read the matching layer notes, check the Source Map, and search the listed symbols in claudecode_src/src/. Do not start by reading the source tree from top to bottom. If you get lost in the source, keep the Source Navigation Guide open next to it.
Use the Study Exercises. They force you to record the symbols you searched, the files you inspected, and the argument you can defend from the code.