Boost Development Efficiency with Claude Code: 16 Essential Commands and a Proven Workflow

Efficiency Comparison

Traditional manual approaches (hand‑flipping directories, accumulating noisy context, reusing old conversations) are contrasted with Claude Code’s command‑driven workflow. Using /init for project onboarding, /compact to compress long dialogs, /clear to start fresh, and Plan Mode for pre‑planning consistently improves speed and reduces errors.

Four Core Commands

The most frequently used commands are /init, /compact, /clear, and /memory. Each command is described with its purpose and the scenarios where it shines.

Practical example: when taking over an unfamiliar repository, run /init then /memory to add rules such as “do not touch legacy directory” or “test command is pnpm test ”, saving repeated explanations.

Three Interaction Modes

Claude Code offers Default, Auto‑Accept, and Plan Mode. Default confirms each step, Auto‑Accept runs file changes automatically, and Plan Mode only analyzes without modifying code. Each mode is matched to suitable tasks.

Use case: adding OAuth support—first run Plan Mode to map the call chain and split change points, then switch to an execution mode to implement with low rework risk.

16 High‑Frequency Commands

A concise table lists each command, its function, and the optimal timing, e.g., claude --version for version checks, claude to start in the current directory, /status to inspect session state, /cost to monitor token usage, and shortcut keys Ctrl+B and Ctrl+T for background tasks and task panel access.

Why /init Is Critical

Running /init makes Claude Code scan the repository and generate a CLAUDE.md draft. Adding project‑specific rules, code‑style guidelines, forbidden directories, and test commands to this file makes subsequent interactions more accurate.

Example: in an existing project, run /init then append “commit messages must follow feat/fix/chore” and “do not modify src/legacy/ ”. Future outputs then respect these conventions.

/compact vs /clear

Both commands tidy context but serve different purposes. /compact compresses an ongoing conversation while preserving key information; /clear wipes the history for a brand‑new task.

Use /compact to continue the same task when the dialog becomes too long.

Use /clear when switching to a completely different task.

Scenario: after finishing a payment module, start the permission system with /clear for a clean slate; if the payment module is unfinished but the dialog is long, use /compact "retain payment flow and unresolved issues" .

/memory – The Hidden Ace

CLAUDE.md

acts as a long‑term project memory. Store technical stack, directory boundaries, common commands, commit standards, and team preferences. The earlier this information is recorded, the more it reduces future friction.

- Do not modify <code>src/legacy/</code>
- API files live in <code>src/api/</code>
- Test command is <code>pnpm test</code>
- Keep responses concise; avoid unnecessary refactoring

Model Selection Logic

Choose the Claude model based on task type: Sonnet for everyday coding and simple changes, Opus for architecture analysis and complex planning, and Haiku for quick Q&A.

Example: use Sonnet for bug fixes, Opus when dissecting a complex call chain, and Haiku for rapid look‑ups.

Practical Workflow

Run /init to let Claude Code understand the project.

Enter Plan Mode to analyze structure and plan changes.

Switch to Default or Auto‑Accept to execute the plan.

Offload long‑running tasks (tests, builds) with Ctrl+B.

Use /compact for long dialogs or /clear when changing tasks.

Write back experience with /memory for future reuse.

Example: for a new feature, start with /init , outline the plan in Plan Mode, execute changes, run tests in the background, then record “OAuth logic belongs in layer X” via /memory for the next iteration.

Common Pitfalls

Five frequent mistakes are highlighted: jumping straight to code changes, letting dialogs become messy, continuing old conversations for new tasks, modifying without prior planning, and not persisting project rules. The recommended stable practices are to use /init, /compact, /clear, Plan Mode, and /memory appropriately.

Conclusion

When the modes, commands, memory, and model choices are orchestrated into a stable workflow, Claude Code transforms from a simple chat assistant into a continuous AI programming partner that reliably boosts development efficiency.

Claude Code official docs: https://docs.anthropic.com/claude-code
Aliyun article: https://developer.aliyun.com/article/1724549