Module 6 of 6

Share Your Skill

10 min • 0 Lessons

Share Your Skill

Just for you, your team, or the world. Three paths, none of them complicated.

Three Ways to Share. Pick the One That Fits.

You have a working skill. It does its job. The voice is yours.

Now what?

There are three answers, depending on who needs it.

Just you
Your team
Or the wider Claude community

We will go through each one, in plain English, with the exact steps.

Option 1: For Just You. You Are Already Done

If the skill is only for you, you have already finished.

The skill folder lives on your computer (if you built it in Claude Code) or on your account (if you uploaded it through web Settings).

It will follow you everywhere.

The only thing worth doing: back it up.

Take the SKILL.md file and drop it somewhere that survives your laptop dying. Three good options:

A Notion page or doc, pinned in a "My Skills" section.
A private GitHub repo called my-claude-skills.
An iCloud / OneDrive folder you already back up.

That is the whole "for just you" workflow. Move on.

Option 2: For Your Team (Two Paths)

If your team uses Claude, your skill is more useful in their hands than in yours alone.

There are two ways to share, depending on which Claude surface your teammates use.

If Your Team Uses Claude Web or Desktop

Step 1: Zip the skill folder.

💻In Claude Code's terminal, run

cd ~/.claude/skills && zip -r meeting-notes-summary.zip meeting-notes-summary/

You now have a .zip file with the whole skill in it.

It is a few kilobytes. Email-able. Slack-able.

Step 2: Send the zip to your teammates.

Slack, email, a shared drive, whatever your team uses.

The file is small enough to attach anywhere.

Step 3: They upload it to Claude.

Each teammate opens claude.ai, goes to Settings → Capabilities → Skills (the exact label depends on plan), clicks Upload Skill, and picks the zip.

That is it. The skill is now in their account.

💡

The "everyone on the team" version. If you are on Claude Enterprise, an admin can deploy a skill workspace-wide. Every account gets it automatically, and updates push to everyone at once. Ask your admin. The DIY zip-and-send works on any plan.

If Your Team Uses Claude Code

Drop the folder into their ~/.claude/skills/ directory.

They restart Claude Code. Done.

The cleanest way is a shared git repo:

💻First time setup (one teammate does this once)

cd ~/.claude/skills
git init
git remote add origin git@github.com:yourcompany/team-skills.git
git add meeting-notes-summary
git commit -m "Add meeting-notes-summary skill"
git push -u origin main

💻Every other teammate runs this once

cd ~/.claude
git clone git@github.com:yourcompany/team-skills.git skills

After that, anyone who improves a skill pushes the change.

Anyone who needs the latest version pulls.

Skills become a normal part of your team's git workflow.

💡

Why git is better than a shared drive for skills. You get version history, you can see who changed what, and you can revert if a change makes things worse. Skills are just text. They live in git beautifully.

Option 3: For the World. Publish on GitHub

If you built a skill that solves a generic problem (not "for my company") you can publish it.

Other people install it. Some of them improve it and submit fixes back.

This is roughly the same as publishing any open-source project.

Step 1: Make a public GitHub repo.

Name it something descriptive: meeting-notes-summary-skill or claude-skill-meeting-notes.

Public visibility.

Step 2: Put your skill folder in the repo.

Drop the meeting-notes-summary/ folder at the root of the repo.

Don't put a README.md inside the skill folder (that confuses Claude).

Put the README.md at the repo root, outside the skill folder.

Step 3: Write the README for humans, not Claude.

The README is the page someone sees when they land on your repo. It should explain:

What the skill does, in one sentence.
Who it is for.
How to install it (one paragraph: drop the folder into ~/.claude/skills/ or upload via Settings).
A quick before/after example so people know what they are getting.
A screenshot or two of it in action.

💡

Submit to Anthropic's directory. If your skill is solid, you can submit it to github.com/anthropics/skills as a pull request. They review and merge community skills. Inclusion gives you discoverability.

What We Are Deliberately Not Covering

Some skills are built into bigger applications using the Skills API and Anthropic's Agent SDK.

That is real, that is useful, and it is genuinely a different course.

If you are a developer building an app that uses skills programmatically, the Anthropic Skills API docs are where you go next.

Everything else in this course (the descriptions, the trigger fixes, the iteration loop) still applies, the only thing that changes is how the skill gets loaded.

For the rest of you: you have shipped a real skill.

You can use them, build them, debug them, and share them. That is the whole beginner course.

What You Have Done in About an Hour

Take stock.

You started this course not sure what a Claude Skill was. You finish it having:

Understood why skills exist and when to use them (Module 1).
Learned how skills work across Claude Web, Desktop, and Code (Module 2).
Used a skill someone else built and felt the difference (Module 3).
Built your own first skill with skill-creator (Module 4).
Fixed a skill that did not trigger correctly (Module 5).
Shared a skill with your team or the world (Module 6).

That is the operator-level grasp.

You can now turn any repeated workflow into a permanent capability for Claude, and for anyone you share it with.

✨ What you just did. You moved from "I prompt Claude every time" to "I teach Claude once". That is the operator move. Most people never make it. You did.

Where to Go Next

You have shipped one skill. Here is what gets powerful next.

Chain Skills Together

The real magic shows up when one skill triggers another.

Picture writing a blog post. Today you might do it in three separate Claude conversations:

Research what is hot in your industry this week
Draft a post in your voice
Run an SEO check and fix the gaps

With three skills (research, my-voice-draft, seo-check), you can chain them in a single ask:

💻Type something like

Research what is hot in the [your industry] space this week,
then draft a 1200-word post in my voice on the most interesting
angle, then run the SEO check and tell me what to fix.

Claude loads all three skills, runs them in sequence, and hands you a finished draft.

We use this exact chain at Waboom AI. One ask, three skills, one finished post.

Build Skills That Use Connectors

Skills get even more useful when they pair with Connectors (the access layer that lets Claude read your Notion, send emails, update your CRM, etc).

A skill on its own can only work on what you paste into the chat. A skill plus a connector can:

Read your Linear backlog and draft this sprint's tasks
Read your inbox and draft replies in your voice
Read your HubSpot deal pipeline and write the weekly forecast

When you build your next skill, ask yourself: what would this be able to do if it could reach out to one of my tools? If the answer is "a lot more", that is your second skill.

We cover the full distinction in Claude Skills vs MCP on the blog.

Use Skills Other People Have Built

You do not have to build every skill from scratch. There are three places to find skills built by other people.

The Anthropic skills repo at github.com/anthropics/skills. Official, maintained, safe. Start here.
The Skills section inside Claude Desktop. Click the + button next to Skills and then Browse skills. Anthropic ships a directory of community plugins you can install in one click.
Operator newsletters. Karol Zieminski's Substack ships specific, tested skills (alt-text generator, LinkedIn teasers, expense trackers) with the full SKILL.md inline. Read, copy, adapt.
GitHub at large. Search "claude skill" on GitHub and you will find dozens of community skills. Real estate skills, finance skills, design skills, dev tooling.

When you install a skill from any of these sources, open the SKILL.md file and read it first. Always.

A skill is just instructions Claude follows. If you install one without reading it, you are letting a stranger tell your Claude what to do. Most of them are fine. Some are not.

A 30-second read is the difference between a useful capability and an unexpected behaviour. Treat it like installing a browser extension or a script someone shared on Reddit.

💡

Trust check. Before installing a skill from someone you do not know: (1) read the SKILL.md end to end. (2) Check if there are scripts in a scripts/ folder, those run when triggered. (3) Look at the author's other work. (4) Prefer skills from the official Anthropic repo or from a person you can verify.

✨ The change-log habit. Karol Zieminski's rule for any autonomous work: always ask for a what-changed.md summary after the task. "Always ask for a change log." It is a 10-second audit that catches every silent decision a skill made. Build the habit early, especially with skills you did not write yourself.

🎓Go further: The 5 advanced patterns Anthropic has seen workclick to expand

Once you have built two or three skills, you will start spotting these shapes. Anthropic published them as "patterns we have seen work well" and they are a strong head start for skill number four.

1. Sequential workflow orchestration. Your skill walks Claude through a fixed, ordered set of steps. Step 1 must finish before Step 2 starts. Each step has a check. Good for: customer onboarding flows, sprint setup, invoicing.

2. Multi-MCP coordination. Your skill orchestrates calls across two or more Connectors. Example: pull from Figma, store in Drive, create tickets in Linear, post a summary to Slack. Each phase hands data to the next. Good for: design-to-dev handoff, weekly reports that aggregate from multiple sources.

3. Iterative refinement. The skill knows the output gets better when it loops. Draft → validate → fix → re-validate → polish. The skill defines the quality threshold and knows when to stop iterating. Good for: report generation, complex copy, anything where "draft one" never matches "draft three".

4. Context-aware tool selection. The skill has a decision tree. Same outcome (e.g., store this file), different tool depending on context (cloud storage for large files, Notion for collaborative docs, GitHub for code, local for temporary). Good for: routing decisions, fallback logic.

5. Domain-specific intelligence. The skill embeds expert knowledge a generic LLM would miss. Financial compliance rules. Medical screening rules. Legal contract red flags. The skill enforces the domain rules before any action, including a paper trail. Good for: anything where mistakes have a regulator attached.

You do not need to pick a pattern up-front. Build the skill, and the pattern will reveal itself in the shape your steps take.

What's Next After This Course

A few good directions:

Build more skills. The first one is the hardest. Once you have one, you will start spotting candidates everywhere. "I do this every Monday" is a skill. "I have to remind Claude how to format this" is a skill.
Share your skill with us. Email it to leo@waboom.ai or tag us on LinkedIn. We love seeing what people build.
Go deeper on Claude Code. If Module 4 Path C was your first time in the terminal, our Claude Code 101 course covers the rest of what Claude Code can do.
Come to a workshop. We run live AI Operator workshops every month. Hands-on Claude Code + Skills training. Bring a real problem, leave with a working skill.

Thanks for doing the course.

←PreviousWhy Skills Don't Trigger Course Complete!Back to Resources →

←

Module 6 of 6

Share Your Skill

10 min • 0 Lessons

Share Your Skill

Just for you, your team, or the world. Three paths, none of them complicated.

Three Ways to Share. Pick the One That Fits.

You have a working skill. It does its job. The voice is yours.

Now what?

There are three answers, depending on who needs it.

Just you
Your team
Or the wider Claude community

We will go through each one, in plain English, with the exact steps.

Option 1: For Just You. You Are Already Done

If the skill is only for you, you have already finished.

The skill folder lives on your computer (if you built it in Claude Code) or on your account (if you uploaded it through web Settings).

It will follow you everywhere.

The only thing worth doing: back it up.

Take the SKILL.md file and drop it somewhere that survives your laptop dying. Three good options:

A Notion page or doc, pinned in a "My Skills" section.
A private GitHub repo called my-claude-skills.
An iCloud / OneDrive folder you already back up.

That is the whole "for just you" workflow. Move on.

Option 2: For Your Team (Two Paths)

If your team uses Claude, your skill is more useful in their hands than in yours alone.

There are two ways to share, depending on which Claude surface your teammates use.

If Your Team Uses Claude Web or Desktop

Step 1: Zip the skill folder.

💻In Claude Code's terminal, run

cd ~/.claude/skills && zip -r meeting-notes-summary.zip meeting-notes-summary/

You now have a .zip file with the whole skill in it.

It is a few kilobytes. Email-able. Slack-able.

Step 2: Send the zip to your teammates.

Slack, email, a shared drive, whatever your team uses.

The file is small enough to attach anywhere.

Step 3: They upload it to Claude.

Each teammate opens claude.ai, goes to Settings → Capabilities → Skills (the exact label depends on plan), clicks Upload Skill, and picks the zip.

That is it. The skill is now in their account.

💡

If Your Team Uses Claude Code

Drop the folder into their ~/.claude/skills/ directory.

They restart Claude Code. Done.

The cleanest way is a shared git repo:

💻First time setup (one teammate does this once)

cd ~/.claude/skills
git init
git remote add origin git@github.com:yourcompany/team-skills.git
git add meeting-notes-summary
git commit -m "Add meeting-notes-summary skill"
git push -u origin main

💻Every other teammate runs this once

cd ~/.claude
git clone git@github.com:yourcompany/team-skills.git skills

After that, anyone who improves a skill pushes the change.

Anyone who needs the latest version pulls.

Skills become a normal part of your team's git workflow.

💡

Option 3: For the World. Publish on GitHub

If you built a skill that solves a generic problem (not "for my company") you can publish it.

Other people install it. Some of them improve it and submit fixes back.

This is roughly the same as publishing any open-source project.

Step 1: Make a public GitHub repo.

Name it something descriptive: meeting-notes-summary-skill or claude-skill-meeting-notes.

Public visibility.

Step 2: Put your skill folder in the repo.

Drop the meeting-notes-summary/ folder at the root of the repo.

Don't put a README.md inside the skill folder (that confuses Claude).

Put the README.md at the repo root, outside the skill folder.

Step 3: Write the README for humans, not Claude.

The README is the page someone sees when they land on your repo. It should explain:

What the skill does, in one sentence.
Who it is for.
How to install it (one paragraph: drop the folder into ~/.claude/skills/ or upload via Settings).
A quick before/after example so people know what they are getting.
A screenshot or two of it in action.

💡

What We Are Deliberately Not Covering

Some skills are built into bigger applications using the Skills API and Anthropic's Agent SDK.

That is real, that is useful, and it is genuinely a different course.

If you are a developer building an app that uses skills programmatically, the Anthropic Skills API docs are where you go next.

Everything else in this course (the descriptions, the trigger fixes, the iteration loop) still applies, the only thing that changes is how the skill gets loaded.

For the rest of you: you have shipped a real skill.

You can use them, build them, debug them, and share them. That is the whole beginner course.

What You Have Done in About an Hour

Take stock.

You started this course not sure what a Claude Skill was. You finish it having:

Understood why skills exist and when to use them (Module 1).
Learned how skills work across Claude Web, Desktop, and Code (Module 2).
Used a skill someone else built and felt the difference (Module 3).
Built your own first skill with skill-creator (Module 4).
Fixed a skill that did not trigger correctly (Module 5).
Shared a skill with your team or the world (Module 6).

That is the operator-level grasp.

You can now turn any repeated workflow into a permanent capability for Claude, and for anyone you share it with.

✨ What you just did. You moved from "I prompt Claude every time" to "I teach Claude once". That is the operator move. Most people never make it. You did.

Where to Go Next

You have shipped one skill. Here is what gets powerful next.

Chain Skills Together

The real magic shows up when one skill triggers another.

Picture writing a blog post. Today you might do it in three separate Claude conversations:

Research what is hot in your industry this week
Draft a post in your voice
Run an SEO check and fix the gaps

With three skills (research, my-voice-draft, seo-check), you can chain them in a single ask:

💻Type something like

Research what is hot in the [your industry] space this week,
then draft a 1200-word post in my voice on the most interesting
angle, then run the SEO check and tell me what to fix.

Claude loads all three skills, runs them in sequence, and hands you a finished draft.

We use this exact chain at Waboom AI. One ask, three skills, one finished post.

Build Skills That Use Connectors

Skills get even more useful when they pair with Connectors (the access layer that lets Claude read your Notion, send emails, update your CRM, etc).

A skill on its own can only work on what you paste into the chat. A skill plus a connector can:

Read your Linear backlog and draft this sprint's tasks
Read your inbox and draft replies in your voice
Read your HubSpot deal pipeline and write the weekly forecast

When you build your next skill, ask yourself: what would this be able to do if it could reach out to one of my tools? If the answer is "a lot more", that is your second skill.

We cover the full distinction in Claude Skills vs MCP on the blog.

Use Skills Other People Have Built

You do not have to build every skill from scratch. There are three places to find skills built by other people.

The Anthropic skills repo at github.com/anthropics/skills. Official, maintained, safe. Start here.
The Skills section inside Claude Desktop. Click the + button next to Skills and then Browse skills. Anthropic ships a directory of community plugins you can install in one click.
Operator newsletters. Karol Zieminski's Substack ships specific, tested skills (alt-text generator, LinkedIn teasers, expense trackers) with the full SKILL.md inline. Read, copy, adapt.
GitHub at large. Search "claude skill" on GitHub and you will find dozens of community skills. Real estate skills, finance skills, design skills, dev tooling.

When you install a skill from any of these sources, open the SKILL.md file and read it first. Always.

A skill is just instructions Claude follows. If you install one without reading it, you are letting a stranger tell your Claude what to do. Most of them are fine. Some are not.

A 30-second read is the difference between a useful capability and an unexpected behaviour. Treat it like installing a browser extension or a script someone shared on Reddit.

💡

✨ The change-log habit. Karol Zieminski's rule for any autonomous work: always ask for a what-changed.md summary after the task. "Always ask for a change log." It is a 10-second audit that catches every silent decision a skill made. Build the habit early, especially with skills you did not write yourself.

🎓Go further: The 5 advanced patterns Anthropic has seen workclick to expand

Once you have built two or three skills, you will start spotting these shapes. Anthropic published them as "patterns we have seen work well" and they are a strong head start for skill number four.

You do not need to pick a pattern up-front. Build the skill, and the pattern will reveal itself in the shape your steps take.

What's Next After This Course

A few good directions:

Build more skills. The first one is the hardest. Once you have one, you will start spotting candidates everywhere. "I do this every Monday" is a skill. "I have to remind Claude how to format this" is a skill.
Share your skill with us. Email it to leo@waboom.ai or tag us on LinkedIn. We love seeing what people build.
Go deeper on Claude Code. If Module 4 Path C was your first time in the terminal, our Claude Code 101 course covers the rest of what Claude Code can do.
Come to a workshop. We run live AI Operator workshops every month. Hands-on Claude Code + Skills training. Bring a real problem, leave with a working skill.

Thanks for doing the course.

←PreviousWhy Skills Don't Trigger Course Complete!Back to Resources →