For developers who already know the basics of Claude Code. Learn to use CLAUDE.md, Settings, permissions, Skills, Commands, subagents, Hooks, workflows, and diagnostic tools to turn ad hoc conversations into clear, reusable, and verifiable project collaboration.
Choose project context worth retaining, place rules at the correct scope, and rewrite them as project instructions that Claude Code can follow and verify.
Choose the right Settings file for each sharing scope, explain scalar overrides and array merging, and write a small, clear Claude Code configuration that can be verified.
Predict `deny`, `ask`, and `allow` matches, design least-privilege rules for Bash, files, network access, MCP, and subagents, and verify the configuration against real scenarios.
Decide whether content belongs in a Skill and design a capability package with clear triggers, progressive loading, explicit tool boundaries, and verifiable success and failure paths.
Distinguish the different slash-menu entries and design repeatable, stable workflows as custom commands with clear arguments, stop conditions, and an appropriate executor.
Decide when delegation is worthwhile, give a subagent complete boundaries and minimal tools, and control context, file ownership, and verification during parallel execution.
Choose Hook events, Matchers, and handlers by lifecycle; use inputs, exit codes, synchronous and asynchronous behavior correctly; and diagnose trigger or script failures through logs.
Work backward from acceptance evidence to define research, planning, implementation, verification, and review stages, with inputs, outputs, checkpoints, stop conditions, and suitable execution mechanisms for each stage.
Turn vague failures into reproducible symptoms, choose diagnostic tools across installation, connection, context, configuration, permissions, Hooks, and MCP layers, and verify the fix through the original failure path.
