Claude Desktop Skills: How to Install and Use Them

Claude desktop skills are reusable instruction files that you place in a folder the Claude desktop app reads at startup. Once installed, you invoke a skill with its slash command and Claude follows the instructions in that file. Skills save time on tasks you repeat and make consistent results easier to achieve across sessions.

Installing a skill takes a minute; deciding which skills to keep and how to share the best ones with your team takes more care. Tools like fenbs store approved skill content and team usage notes outside the desktop app, so the record survives an app reinstall or a change of machine.

What Claude desktop skills are

A Claude desktop skill is a markdown file whose name becomes a slash command. When you type /skill-name in the desktop app, Claude reads the file and applies its instructions to the conversation. The skill does not run code; it adds context and steps that shape how Claude responds.

Desktop skills are distinct from Claude Code skills and from claude.ai project instructions. Each surface has its own folder and its own scope. See claude skills files for the full picture of where files live across environments, and claude skills marketplace for finding community skill collections.

How to install a skill in the Claude desktop app

1. Locate the skills folder for the Claude desktop app. On macOS this is typically ~/.claude/skills/; check the app's documentation or settings to confirm the path for your version.
2. Create the folder if it does not exist. Use your file manager or run mkdir -p ~/.claude/skills in a terminal.
3. Copy or create a skill file in that folder. The file must be a markdown document with a .md extension. Name it using lowercase letters and hyphens, for example summarize.md.
4. Restart the Claude desktop app. Skills are read at startup; a file added while the app is running will not appear until you restart.
5. Type / in a new conversation to confirm the skill appears in the list. If it does not, check the filename and folder path.

Writing a skill file for the desktop app

Start the file with a one-sentence description of what the skill does. Follow it with numbered steps. End with a short note on what the skill is not meant for. That last line prevents the skill from being misused on tasks it was never designed to handle.

# Meeting notes
Turn a rough transcript into structured meeting notes.

1. Identify the topic and the date from the input.
2. List decisions made, one per bullet.
3. List action items, each with the person responsible.
4. List open questions that were not resolved.

Not for: editing polished documents or writing minutes for publication.

Testing a skill after installation

Run the skill against a small, real example before relying on it for important work. Use input you already know the right answer for, compare the output to your expectation and note any gaps. A skill that produces inconsistent results on simple cases will not improve on complex ones.

If the skill misses something, edit the file directly and restart the app to reload it. Make changes one at a time. A single edit followed by a retest tells you whether that change helped. Multiple simultaneous edits make it hard to know what fixed or broke the behavior.

Sharing skills with a team

Desktop skills are local by default. If your team shares a project, the most reliable way to share skills is to save the file content in a shared location — a project repository, a shared drive or a team knowledge tool — and have each person install it locally from there.

When a skill changes, the person who made the update needs to notify the rest of the team so they can update their local copies. Without coordination, team members end up with different versions of the same skill, which produces different results on the same task.

When to update a desktop skill

Update a desktop skill when it consistently produces a result that requires correction, when a new task type reveals a gap in the instructions, or when a model update changes how the instructions are interpreted. Edit the file, restart the app and retest before relying on the updated version for real work.

Before overwriting the skill file, save the current version under a different name. If the update makes things worse, you can restore the old instructions immediately without trying to reconstruct what they said. Keeping one previous version is usually enough.

Let teammates know when a shared skill changes. An update that goes out without notice means people end up using different versions of what they believe is the same instruction set. A short note saying what changed and why is enough to keep the team aligned.

How to do this with fenbs

1. Create a free account and add a project for the team that will use these skills.
2. After testing a desktop skill and approving it, paste its instruction text into fenbs under Skills and prompts. Add a short note about what it does and any usage limits.
3. When a team member asks which skills to install, point them to the fenbs record. They copy the text, create the local file and restart the app.
4. Connect your AI app over MCP from Integrations. A connected assistant reads your approved skills at the start of each session and applies them without requiring a local file on every machine.