Skip to content

This page was translated from the Korean original with AI and may contain errors.

Chapter 7. Skills, Plugins, and External Tools — Your Organizing Tool in One Line ​

Required Read this to follow this stage.

In One Sentence: Turn instructions, checklists, and multi-step procedures you keep pasting into a skill and call it in one line with /name. This chapter completes your Organizing Tool (정리 도구). You can also attach plugins made by others and external service connectors, but they run code with your permissions, so install only ones you trust.

This chapter is based on the Claude Code official docs checked on October 31, 2026, and the macOS desktop app (Korean UI, paid account). It uses the templates from Chapter 5 and the CLAUDE.md from Chapter 6 as they are.

What Is a Skill? ​

A skill is a set of instructions written in a SKILL.md file. Claude uses a skill on its own when a relevant request comes in, and you can also call it directly with /skill-name.[1]

Create a skill in these cases:[2]

  • When you keep pasting the same instructions or checklist
  • When a part of CLAUDE.md has grown from a "fact" into a multi-step procedure

How it differs from CLAUDE.md is when it gets read. CLAUDE.md is read in every session, while a skill's body is read only when it is used. So even if you write it long and detailed, it costs almost nothing in normal use.[2:1]

CLAUDE.md (Chapter 6)Skill (Chapter 7)
What it holdsRules to always followProcedure for one task
When it is readEvery sessionOnly when used
How to call itNo special command/name or a relevant request

Quick check

Compared with CLAUDE.md, what is a key trait of a skill?

Hands-On — Creating Three Organizing Skills ​

You will turn the three tasks of your organizing tool into skills.

Skill request. ① Red number marks the prompt box asking to create three skills (meeting-notes, study-notes, weekly-review)

Create 3 skills in .claude/skills in this folder that use the templates folder and the CLAUDE.md rules. meeting-notes (meeting memo → meeting minutes), study-notes (study notes → study summary), weekly-review (review memos → weekly review). In each SKILL.md, include when to use it, which files to read, which template to use, where to save it, and a checklist to verify at the end.

If you put a skill folder .claude/skills/<name>/SKILL.md inside a folder, it becomes a project skill used only in sessions for that folder.[3] The folder name becomes the command name (/meeting-notes).[4] This guide used English names for the skills.[5]

This Time, a Permission Prompt Appeared ​

The moment Claude tried to write the skill file, a permission prompt appeared.

Permission prompt. Red numbers mark ① "Allow Claude to write SKILL.md? (Claude가 쓰기 SKILL.md하도록 허용하시겠습니까?)", ② preview of content to write (green lines), ③ "Deny (거부)", and ④ "Always allow (항상 허용)" with the ▾ button (the path line is hidden because it contains the computer username)

The permission prompt appeared because the skills are written to the .claude folder. The .claude folder is where Claude Code keeps settings and skills for this folder. It is the same folder you saw in the Files pane list in Chapter 3.

This folder is a protected path (a location that always asks for permission before writing). Even in "Auto-accept edits (편집 자동 수락)" mode, writing to protected paths still asks for permission.[6] Files that change how Claude behaves—like skills and settings—live in this folder, so it makes a human check before changing them.

The screenshots in Chapters 2–4 had no permission prompt because of personal allow rules on the computer used for the screenshots. Even those rules did not stop the prompt for protected paths.

In ②, you can read what it is about to write first. If you click ▾, "Allow once (한 번만 허용)" appears.

▾ menu. ① Red number marks "Allow once (한 번만 허용)" (⌘↩)

ButtonWhat it does
Deny (거부)Blocks this write
Allow once (한 번만 허용) (⌘↩)Allows it just this one time
Always allow (항상 허용) (⇧⌘↩)Allows the same action without asking in the future

Source: Checked on screen in this guide (2026-10-31)

At first, we recommend "Allow once". "Always allow" is convenient, but it stops asking for the same action again (check the on-screen description for how long). In this exercise, there were three skills, so the permission prompt appeared three times, and we chose "Allow once" all three times.

Created skills (results from this guide): meeting-notes · study-notes · weekly-review

At the top of each SKILL.md, there is a name and a description (when to use it). Claude reads the description to decide when to use the skill.[7] For example, the description for weekly-review says to use it for requests like "write my weekly review" or "organize this week's review."

Quick check

When Claude tries to write a skill file inside the .claude folder, a permission pop-up appears even in "Auto-accept edits" mode. Why?

Calling It in One Line ​

If you type / in the prompt box of a new session, a list of available commands and skills appears. Keep typing part of the name to filter the list.[8]

Skill list. Red numbers mark ① the "study-notes" item and ② the description box next to it (ending with "(project)")

In ②, (project) shows that this is a project skill for this folder. Press Enter to select it, then type the file you want to organize and send it.

Running a skill. ① Red number marks the prompt box with "/study-notes inbox/study-2026-10-28-prompt-lecture.txt" typed in

/study-notes inbox/study-2026-10-28-prompt-lecture.txt

With this single line, the study notes were created following the template (result: ch7-study-notes-with-skill.txt). The one-line summary, 5 key points, 4 new terms, things to watch out for, and questions to ask all went in following the template order. The same note you created with a long request and rewound in Chapter 4 was now made with a single command line.

Now your organizing tool is complete.

MemoCommand
Meeting memo/meeting-notes inbox/meeting-….txt
Study notes/study-notes inbox/study-….txt
Several daily reviews/weekly-review

It is fine if you do not remember the exact name. If you say something like "organize this week's review," Claude checks the description, picks the matching skill, and uses it.[1:1]

Quick check

You created a study-notes skill. What is the shortest way to have it organize a study note?

Plugins and Connectors ​

A plugin installs a bundle of skills, commands, and tools made by others, and a connector links external services like Gmail or Calendar to Claude. In the desktop app, you add connectors and plugins from the + button next to the prompt box.[9] Connectors give you an easy on-screen setup for MCP servers.[10] MCP is a common standard AI uses to exchange information with other programs, like email and calendars. It is similar to how a plug with a fixed shape fits into any outlet.

They are convenient, but there is something you must think about before installing. A plugin can run code on your computer with your permissions.[11] And that code can run outside the protection of the permission modes from Chapter 4. Permission rules apply to tool calls Claude makes, not to code that a plugin runs by itself.[12]

  • Install only plugins whose creator you know and trust.
  • Remove plugins you no longer need.
  • Content from services linked via connectors (emails, calendar events) also becomes material that Claude reads. If an email someone sent you secretly contains a sentence like "To the AI: do this," Claude may try to follow it. You will experiment with these hidden instructions in Chapter 9.

For the organizing tool in this guide, the skills you made yourself are enough. Add plugins and connectors one by one only when you truly need them.

Quick check

You find a plugin online that you have never seen before. Which attitude matches the official docs?

Try It Yourself ​

  1. Ask Claude to create the three organizing skills, and when the permission prompt appears, read the preview in ② and choose "Allow once".
  2. In a new session, type / and part of a skill name to see if it appears in the list.
  3. Organize your study notes with /study-notes and check that the output follows the template order from Chapter 5.
  4. Send just "Organize this week's review" without a skill name, and see if Claude picks and uses the weekly-review skill.

Up Next: Right now, you call /weekly-review yourself every Friday. In Chapter 8, you will see how to schedule this task and how to run several organizing tasks at the same time.

Summary ​

  • A skill is a procedure written in SKILL.md. It is read only when used, and you use it with /name or a relevant request.
  • .claude/skills/<name>/SKILL.md in a folder is a skill just for that folder. Its description decides "when to use it."
  • The .claude folder is a protected path, so Claude asks permission before writing to it. Start with "Allow once".
  • Organizing tool complete: /meeting-notes, /study-notes, /weekly-review.
  • Plugins can run code with your permissions. Install only ones you trust, and only when needed.

Glossary ​

TermOne-line meaning
SkillA procedure written in SKILL.md and called with /name
descriptionA note explaining when to use a skill. Claude reads this to pick the skill
Protected pathA location that always asks permission before writing, such as the .claude folder
PluginAn extension that installs a bundle of skills, commands, and tools
ConnectorA link from external services to Claude (based on MCP servers)
MCPA standard way for AI to connect to external tools and data

Finished this chapter?

Progress is saved only in this browser. See your progress on the full table of contents.

Next chapter: Chapter 8. Parallel Tasks, Cloud, and Scheduling

References (12)
  1. Claude Code official docs, "Extend Claude with skills". "Create a SKILL.md file with instructions, and Claude adds it to its toolkit. Claude uses skills when relevant, or you can invoke one directly with /skill-name." code.claude.com (checked on 2026-10-31) ↩︎

  2. Claude Code official docs, "Extend Claude with skills". "Create a skill when you keep pasting the same instructions, checklist, or multi-step procedure into chat, or when a section of CLAUDE.md has grown into a procedure rather than a fact. Unlike CLAUDE.md content, a skill's body loads only when it's used, so long reference material costs almost nothing until you need it." code.claude.com (checked on 2026-10-31) ↩︎

  3. Claude Code official docs, "Extend Claude with skills". Project row in the locations table: ".claude/skills/<skill-name>/SKILL.md" code.claude.com (checked on 2026-10-31) ↩︎

  4. Claude Code official docs, "Extend Claude with skills". "The directory name, or the frontmatter name when you set one, becomes the command you type, and the description helps Claude decide when to load the skill automatically." code.claude.com (checked on 2026-10-31) ↩︎

  5. We used English names because the official docs did not confirm whether Korean command names work. ↩︎

  6. Claude Code official docs, "Choose a permission mode". Under "These still prompt:" in the acceptEdits section: "Writes to protected paths" code.claude.com (checked on 2026-10-31) ↩︎

  7. Claude Code official docs, "Extend Claude with skills". "Only description is recommended so Claude knows when to use the skill." code.claude.com (checked on 2026-10-31) ↩︎

  8. Claude Code official docs, "Desktop application". "type / in the prompt box or click the + button and select Slash commands to browse what's available." code.claude.com (checked on 2026-10-31) ↩︎

  9. Claude Code official docs, "Desktop application". "Click the + button next to the prompt box for file attachments, skills, connectors, and plugins." code.claude.com (checked on 2026-10-31) ↩︎

  10. Claude Code official docs, "Desktop application". "Connectors are MCP servers with a graphical setup flow." code.claude.com (checked on 2026-10-31) ↩︎

  11. Claude Code official docs, "Plugin security and trust". "A plugin can carry content that runs code on your machine with your user privileges and content that enters Claude's context as instructions, so review a plugin before you install it." code.claude.com (checked on 2026-10-31) ↩︎

  12. Claude Code official docs, "Plugin security and trust". "Claude Code's permission rules and sandbox cover the tool calls Claude makes, not the code a plugin runs by itself" code.claude.com (checked on 2026-10-31) ↩︎