Guides for Claude Code

Short pages you can read in a few minutes. Each one is written from Anthropic's own documentation, says what it did not test, and links to the source.

• Ten checks for your CLAUDE.mdA short list to review the file Claude Code reads every session.
• Stop an agent from loopingThe limits you can really set, and the ones you cannot.
• A 40-line handoff noteSo a new session picks up where the last one stopped.
• Make Claude Code ask firstQuestions and a short plan before any code is written.
• How to write a SKILL.mdWhere it goes, what to put in it and how to check it loads.
• A starter list of permission rulesAllow, ask and deny, where each file goes, and what rules do not do.
• Three hooks to pasteA notification, formatting after edits and a block on protected files.
• CLAUDE.md, skill, hook or subagentWhich one to use, what each costs in context, and which are enforced.
• Write your first subagentWhere the file goes, the fields that matter and an example to paste.
• Spend less of your quotaWhere tokens go, why long sessions cost more, and the habits that help.
• Which settings file winsThe four places settings come from, the order they win in, and how to check what loaded.
• Add your first MCP serverOne command, the three scopes, the .mcp.json file, and what to check before you trust one.
• Undo changes with /rewindHow checkpoints work, what each restore option does, and what rewind cannot undo.
• Permission modes, explainedThe six modes, how to switch, how plan mode works, and why bypass is for isolated machines.
• Run Claude Code from a scriptclaude -p: piping text in, JSON output, which tools to allow, bare mode, and a warning about repositories you did not write.
• The .claude folder, explainedWhat each file does, which ones to commit, and what Claude Code stores on your disk and for how long.
• Keyboard shortcuts worth learningThe shortcuts that save the most time: stop, undo, switch modes, write longer prompts, search history and edit text.
• How Claude Code remembers your projectCLAUDE.md versus auto memory: what each is for, where it lives, and what to check when Claude ignores it.
• What survives /compactWhat is in the context window, what happens to each kind of content when a session compacts, and what to do before it fills up.
• Parallel sessions with git worktreesGive each Claude Code session its own copy of your files: the --worktree flag, what is shared, copying a .env, and cleanup.
• Customize the status lineSet up a status bar that shows the model, context usage or git branch: the /statusline command, a minimal script, and what to check when it stays blank.
• Debug your configurationWhen Claude ignores your CLAUDE.md, hooks, MCP servers or skills: see what actually loaded, check the usual causes, and test against a clean setup.
• Output stylesChange the tone, length and role of Claude's answers for a whole session: the four built-in styles, how to switch, and how to write your own.
• How Claude Code worksThe loop it runs, the tools it uses, what it can see, how sessions and the context window behave, and the safety nets: a plain-words tour with links to the longer guides.
• Best practices, the short versionTell Claude what it can't see, give it a check to run, enforce the rules that can't have exceptions, and know what to stop doing.
• The sandboxWhat the sandbox covers and what it doesn't, how to turn it on with /sandbox, the two modes, how to widen or narrow it, and its limits.
• Install and manage pluginsHow to install a plugin, choose user, project or local scope, add a marketplace, keep plugins updated, remove them, and see what they cost your context.
• Fix a slow or stuck sessionHigh CPU or memory, freezes, compaction that keeps failing, garbled text, an empty clipboard and searches that find nothing: the fixes, in plain words.
• Set up your terminalShift+Enter for new lines, Option-key shortcuts on macOS, a bell when Claude finishes, tmux, flicker, big pastes and Vim keys: the terminal fixes, in plain words.
• Customize keyboard shortcutsHow to change shortcuts with keybindings.json: the file, contexts, key syntax, chords, unbinding a default, what you can't change and how to catch mistakes.
• Everyday workflowsPrompts and habits for daily work: learn a codebase, fix a bug, refactor, write tests, open a PR, use images and @ references, resume, plan first and schedule runs.
• Run Claude on a schedule with /loopFixed or self-chosen intervals, a default loop.md, one-time reminders, how to stop a loop, the seven-day expiry, jitter, cron syntax and the limits.
• Work toward a goal with /goalSet a completion condition, write one a checker can verify, bound it with a turn limit, check status, clear or resume it, and what stops or pauses a goal.
• Fix “command not found: claude”The install finished but the terminal can't find claude: check the file, put its folder on your PATH on macOS, Linux and Windows, open a fresh terminal, spot duplicate installs and restore a missing claude.exe.
• Glossary in plain wordsThe terms the Claude Code docs use, explained in a sentence or two: agentic loop, context window, compaction, CLAUDE.md, skill, hook, subagent, MCP, permission mode, worktree, cloud session and more.
• Fullscreen mode: flicker, scrolling, copyingSwitch fullscreen rendering on or off with /tui, what changes for search, scrolling and copying, native text selection, tmux and Windows Terminal quirks, and what the startup fallback message means.
• Slash commands worth knowingThe Claude Code slash commands that earn their keep, grouped by the moment you need them: first session, during a task, long conversations, side questions, parallel work, before shipping, between sessions and when something breaks.
• Claude Code in VS CodeInstall the extension, open the panel, sign in, give Claude the right files with @-mentions, review edits, resume or rewind a conversation, the shortcuts worth knowing, extension versus CLI, and what to try when the Spark icon is missing.
• Claude Code in a monorepo or large codebaseKeep Claude focused in a big repository: where to start it, layered CLAUDE.md files, excluding the ones you never need, blocking reads of generated and vendored code, a language-server plugin, sparse worktrees and planning changes that span packages.
• Claude Code in GitHub ActionsQuick and manual setup of the Claude Code GitHub Action, the two secrets and where they go, @claude mentions versus automation with a prompt, who is allowed to trigger it, keeping costs down, the usual fixes when it does not answer, and uninstalling cleanly.
• Claude Code sessions: resume, name and branchContinue the last conversation, pick one from a list, resume by name, name sessions so you can find them, branch a conversation to try another approach, what a resumed session does and does not restore, and where transcripts are kept.
• Claude Code in JetBrains IDEsInstall the CLI and the plugin for IntelliJ, PyCharm, WebStorm and the rest, run claude in the IDE terminal or connect with /ide, the diff viewer and shortcuts, the fixes for ESC, WSL and remote development, and a safety note about edit permissions.
• Claude Code voice dictationSpeak your prompts in the terminal: the requirements and the privacy point (audio goes to Anthropic's servers), /voice, hold mode versus tap mode, cancelling a recording, the dictation language, rebinding the key and the usual fixes when nothing records.
• One skill folder, four tools: SKILL.md in Codex, Cursor and Gemini CLIWhere Claude Code, Codex, Cursor and Gemini CLI look for skills, the .agents/skills shortcut, how to check a skill loaded and call it, and the traps: names, duplicates and consent prompts.

The same guides in Spanish are in Guías. We are not affiliated with Anthropic; the official documentation is the authority.