Getting Started with Claude Code

Claude Code is a terminal-based coding agent. You give it a task, it inspects the repository, edits files, runs commands, and reports what changed. The better mental model is "a coding agent in your shell," not "chat inside a terminal window."

#Basic setup

Before you use it on real work, make sure you have:

• access to the repository
• a working local environment
• a clear brief with scope and constraints

#Good first tasks

Start with tasks that reveal whether the agent can operate inside a real codebase:

• fix a small bug
• rename a shared type
• add i18n support to one component
• update a failing test

Avoid giant greenfield prompts at the start. They make the output look impressive, but they do not tell you whether the agent can actually navigate the repo.

#Prompting habits that help

• define the exact goal
• name the relevant files or modules when possible
• say what must not change
• require verification steps

Example:

txt复制
Add English support to the wiki page shell in jr-academy-web-zh. Keep locale routing intact. Run the narrowest verification that covers the changed files.

#Best working style

Claude Code performs best when you treat it like an engineering collaborator:

• assign a concrete task
• review the implementation
• redirect quickly if the scope drifts

It responds far better to precise constraints than to vague ambition.