Shift log: Claude Code command time-savers, ranked

The commands that save the most time in Claude Code are not necessarily the ones that generate code. They are the ones that prevent repeated explanations, misplaced context, and unnecessary restarts. This shift note compares commands by the operational problem they solve, then names a winner for each situation. Command availability and interface details can change between releases, so check /help in your installed version before building a workflow around a particular command.

/compact beats /clear when the job is still running

Both commands address an overloaded conversation, but they make different trade-offs. /compact summarizes the conversation to reduce its context footprint. /clear removes the conversation context so you can begin again.

Choose /compact when you are still solving the same problem and earlier decisions matter. During a long debugging session, you can supply a focus instruction asking it to preserve the failing test, rejected hypotheses, and relevant file paths. That is more useful than carrying every exploratory exchange forward.

Choose /clear when the task has genuinely changed. A fresh documentation request should not inherit assumptions from an unrelated authentication investigation. Clearing the conversation does not undo file changes, however; inspect your working tree separately.

The verdict is /compact for continuity and /clear for separation. Neither replaces an explicit handoff note when exact details matter. A summary can omit something you later need.

/resume beats reconstructing yesterday’s investigation

A fresh conversation and a resumed session can both get work moving, but only one avoids rebuilding the investigation from memory. /resume lets you return to a previous conversation rather than manually re-entering its history.

Use it when you paused a review, switched repositories, or handed attention to another incident. Before proceeding, tell Claude Code what changed outside that session: new commits, updated dependencies, or a different reproduction case. Restored conversation history is not proof that its assumptions still match the repository.

A fresh session is preferable when the old discussion contains a fundamentally wrong approach. Recovering the transcript can also recover its confusion.

The verdict is /resume for interrupted work and a fresh session for invalidated work. For engineers who frequently switch tasks, session recovery is often the first habit worth adopting.

/memory beats repeated /init runs after onboarding

/init and /memory operate around persistent project guidance, but they belong at different stages. /init helps establish project instructions in CLAUDE.md. /memory provides access to memory files you can maintain as conventions become clearer.

Run /init when entering a repository that lacks useful guidance. Review the result rather than assuming generated instructions are authoritative. Build commands, test expectations, and architectural boundaries should match the actual project.

After that, use /memory to maintain concise, durable instructions. Record a verified test command or an important directory convention, not a transcript of every troubleshooting attempt. Keep secrets out of these files and review shared instruction changes like other repository changes.

The verdict is /init for setup and /memory for ongoing time savings. Repeated initialization is not a substitute for maintaining accurate instructions.

/context beats /cost for diagnosing a crowded session

These commands answer different diagnostic questions. /context helps inspect context usage. /cost reports session usage information, with billing relevance depending on your account and access method.

When a session feels overloaded, start with /context. Its breakdown can help you distinguish conversation growth from other context consumers before deciding to compact or restart. It does not prove why a particular answer was poor.

When the question concerns spending rather than context, use /cost and reconcile it with your provider’s records. Do not treat a local display as a universal billing ledger.

The verdict is /context for context pressure and /cost for usage inspection. Diagnose first; changing models is not automatically the right fix.

The shift handoff prioritizes recovery over command collecting

• Use /compact during a continuing investigation and /clear when crossing into unrelated work.
• Use /resume before spending time reconstructing a valid earlier session.
• Establish project guidance with /init, then maintain it through /memory.

If your workflow also spans several model families and tools, Relay365 is a pay-as-you-go relay desk that works with Claude Code, Codex, Gemini CLI, and OpenAI SDK clients. Its API base is https://api.api365relay.com; OpenAI clients add /v1. Follow its tool-specific configuration instructions rather than assuming every client uses the same setup. Create a Relay365 account if that shared access fits your workflow.

The final verdict is straightforward: preserve useful state, discard irrelevant state, and inspect the actual constraint before changing anything else.