How to Let Claude Code Finish Your Work While You Sleep: 6 Essential Commands

Claude Code (v2.1.154+) provides a complete set of built‑in slash commands that turn a conversational programming assistant into an autonomous worker. By mastering these six core commands you can shift from a "half‑human, half‑Claude" workflow to one where you set the direction and Claude completes the tasks.

/goal — Define a verifiable end condition

Function: Set a goal with a concrete success condition; Claude keeps working until the condition is met, eliminating the need for step‑by‑step prompts.

Problem solved: Without /goal Claude stops after each sub‑task, requiring you to repeatedly say "continue" for multi‑step jobs such as API migration or bulk issue handling.

Mechanism: After each round Claude runs a lightweight model to check the condition; if unmet it proceeds, if met it clears the goal and returns control.

When to use:

Module migration: rewrite all calls from an old API until compilation and tests pass.

Batch fixes: process an issue list until it is empty.

Any multi‑step work with a clear pass/fail criterion.

Best practices: Write verifiable conditions (e.g., "test pass rate 100%"), request evidence (test output, screenshots), monitor progress with /goal (no args), and abort early by re‑issuing /goal if the direction changes.

/loop — Scheduled polling so you can sleep

Function: Re‑run a prompt at a configured interval within the CLI session, useful for status polling or timed reminders.

Problem solved: Long‑running deployments, CI builds, or PR reviews often require you to stare at the terminal; /loop automates periodic checks.

Official description: "Scheduled tasks let Claude re‑run a prompt automatically on an interval. Use them to poll a deployment, babysit a PR, check back on a long‑running build, or remind yourself to do something later in the session."

When to use:

Deployment polling every N minutes.

Monitoring long‑running builds or tests.

Timed personal reminders.

Combine with /goal: set a goal first, then use /loop to monitor progress.

Best practices: Place a loop.md file in the project root to supply a default prompt, choose a sensible interval (5–15 minutes) to avoid token waste or stale information, and switch to cloud‑based Routines for cross‑session persistence.

/batch — Split a large change into parallel tasks

Function: Break a massive codebase change into independent units, each executed in its own Git worktree.

Problem solved: Large‑scale migrations (framework upgrades, dependency version bumps) can overflow Claude's context window or cause conflicts if attempted in one go.

When to use:

Framework migration, e.g., /batch migrate src/ from Solid to React.

Bulk rename or refactor across files.

Standardising code style or dependency versions across modules.

Best practices: Ensure tasks are independent; if task B depends on A's output, prefer /goal. Explicitly specify the directory or file range to avoid accidental changes, and finish with /simplify to clean up.

/simplify — Clean up code after changes

Function: Detect cleanup opportunities in changed code and apply fixes automatically via four parallel review agents (reuse tools, simplify code, optimise efficiency, validate abstraction levels).

Problem solved: Developers often forget to tidy up after a change, leading to technical debt.

Best practices: Run after a feature or fix, before PR submission; target specific paths with /simplify [path]; note that from v2.1.154 onward /simplify no longer performs bug detection—use /code-review for that. Prior to v2.1.154 it behaved like /code-review --fix.

/doctor — Configuration health check

Function: Diagnose core configuration files (CLAUDE.md, settings, hooks, MCP servers, skills) for missing keys, schema errors, or loading failures.

Problem solved: When Claude ignores commands or fails to load skills, the cause is often a mis‑configured file or permission issue.

When to use:

Claude ignores your CLAUDE.md directives.

MCP server connection problems.

Skills not taking effect.

Post‑install configuration verification.

After changing config, confirm correct loading.

Best practices: Run /doctor first, then /context to view what actually made it into the session; this combo quickly pinpoints issues.

/debug — Runtime diagnostics

Function: Enable debug logging so Claude can use logs and configuration paths to diagnose runtime anomalies.

Problem solved: Issues where commands produce wrong results, toolchains are broken, or path resolution fails.

When to use:

Unexpected command behavior.

Toolchain path problems (missing Node, wrong Python version).

Claude Code itself behaving incorrectly.

Before filing a bug report ( /feedback).

Best practices: Run /doctor first to ensure configuration health, then /debug [issue] with a detailed description (e.g., "npm test fails because jest command not found"); after confirming a Claude bug, use /feedback which automatically attaches the debug log.

Workflow composition suggestions

These commands shine when combined. Example snippets:

/goal: Migrate all components from Vue2 to Vue3, compile successfully and achieve test coverage >90%
 → Claude executes autonomously, checking the condition each round

/batch migrate src/components/ from Vue2 to Vue3
 → Parallel large‑scale migration

/simplify src/components/
 → Automatic post‑migration code cleanup

Another pattern:

/goal: Deploy to staging and pass health checks
 → Define the end state

/loop: Check deployment status every 5 minutes, notify when complete
 → Scheduled polling lets you work on other tasks

Diagnostic flow:

Step 1: /doctor          → Config health check (CLAUDE.md, MCP, settings)
Step 2: /context         → Inspect loaded context (system prompt, skills, tools)
Step 3: /debug [issue]  → Enable detailed logs to locate runtime problems
Step 4: /feedback        → Submit bug report if the issue is a Claude bug

References

Claude Code official command docs: https://code.claude.com/docs/en/commands /goal details: https://code.claude.com/docs/en/goal

Scheduled tasks ( /loop) docs: https://code.claude.com/docs/en/scheduled-tasks

Configuration debugging ( /doctor) docs: https://code.claude.com/docs/en/debug-your-config

Claude Code best practices: https://code.claude.com/docs/en/best-practices