How to Install Claude Skills: Step-by-Step Guide (Claude Code & Cursor)

What You Need Before Installing

Before you install any Claude skill collection, you need a working Claude Code installation. Claude Code is Anthropic's terminal-based AI assistant that reads skill files from your project directory. If you haven't installed Claude Code yet, start at the Anthropic documentation at docs.anthropic.com/claude-code and follow the installation guide for your operating system. Claude Code requires a Claude Pro subscription or API key to function.

You also need git installed on your machine if you plan to use the GitHub clone method (the most common and recommended approach). Git is available for free at git-scm.com and comes pre-installed on most macOS systems. You can verify you have git by opening a terminal and running 'git --version' — if you see a version number, you're ready. Windows users should install Git for Windows, which includes the Git Bash terminal.

It's helpful to have a project directory ready — a folder where your actual work files live, or at minimum a dedicated folder you'll use for Claude Code sessions. Claude Code loads skill files from subdirectories within the current working directory, so skills need to be placed within the project context you're working in. If you do all your Claude work in a single directory, you can put skills there and they'll always be available. If you have multiple projects, you'll want to install relevant skills in each project's directory.

Finally, identify which skill collection you want to install. Browse the GTM Skills Directory to find collections that match your primary use case. Look at the collection's GitHub README to confirm it's actively maintained and compatible with your version of Claude Code. Most good collections will list their compatible agents and any prerequisites in the README file.

Method 1: Install via GitHub Clone

GitHub clone is the recommended installation method because it gives you the full collection, makes it easy to get updates, and lets you track your customizations with git. Open your terminal, navigate to your project directory using 'cd path/to/your/project', then run the clone command. For most skill collections, the command follows this pattern: 'git clone https://github.com/[author]/[repo] .claude/skills/[collection-name]'. For example, to install the Agent GTM Skills collection, you would run something like: 'git clone https://github.com/vibeprospecting/agent-gtm-skills .claude/skills/gtm'.

Once the clone completes, list the contents of the installed directory to verify the files are there: 'ls .claude/skills/gtm/' should show you a list of .md files. Each file is a skill covering a specific workflow. You don't need to activate or configure anything — Claude Code will automatically detect and read skill files in the .claude/skills/ directory when it starts a session.

Restart your Claude Code session after installing a new skill collection. Open a new terminal session, navigate to your project directory, and launch Claude Code with 'claude' (or however you start it). At the beginning of the session, Claude Code scans for skill files and loads them into context. You can verify the skills are loaded by asking Claude directly: 'What skills do you currently have loaded?' — Claude should be able to describe the skill files it has access to.

To get updates to a skill collection in the future, navigate to the installed directory and run 'git pull': 'cd .claude/skills/gtm && git pull'. This updates the skill files to the latest version from the source repository. If you've customized any files, git will alert you to conflicts so you can decide how to merge your changes with the upstream updates — one of the key advantages of git-based skill management over manual file copying.

Method 2: Claude Code Marketplace

Claude Code has a built-in skills marketplace that lets you browse and install skill collections without leaving the Claude Code interface. To access it, start a Claude Code session and type '/skills' or access the skills menu from the Claude Code command palette (Ctrl+Shift+P or Cmd+Shift+P on Mac). The marketplace shows available skill collections with descriptions, ratings, and installation instructions.

Browsing the marketplace is intuitive — you can filter by category (sales, marketing, operations, etc.), see how many users have installed each collection, and read the full description before installing. When you find a collection you want, click 'Install' or type the install command shown in the collection's listing. The marketplace handles the file download and directory placement automatically, so you don't need to manually specify where files go.

The marketplace method is the best choice for non-technical users who aren't comfortable with terminal commands or git. The tradeoff is that marketplace-installed collections are slightly less customizable than git-cloned collections — you don't have a local git repository for the skill files, so you can't track your customizations or pull upstream updates as easily. If you later decide you want to customize a marketplace-installed collection extensively, you can convert it to a git-managed installation by initializing a git repo in the installed directory.

Note that marketplace availability varies by Claude Code version and region. If a collection you found in the GTM Skills Directory isn't available in the marketplace, use the GitHub clone method instead. The GTM Skills Directory links directly to each collection's GitHub repository, making the clone method easy regardless of marketplace availability.

Method 3: Manual File Copy

The manual file copy method is the most flexible and works in any environment where you can create and edit files. It's particularly useful when you want to hand-pick specific skill files from a collection rather than installing the entire collection, or when you're working in an environment where git and the marketplace aren't available (like some corporate environments with restricted network access).

To install manually, start by visiting the skill collection's GitHub repository in your browser. You can browse the individual .md skill files directly on GitHub. Click on the file you want, then click the 'Raw' button to see the plain text content. Select all the text (Ctrl+A), copy it (Ctrl+C), and paste it into a new file in your project's .claude/skills/ directory. Name the file with the same .md extension (e.g., 'outbound-research.md').

You can also download individual files from GitHub by clicking the download icon on the raw file page, or by using a wget command in your terminal: 'wget https://raw.githubusercontent.com/[author]/[repo]/main/skills/[file].md -O .claude/skills/[file].md'. This is slightly faster than copy-paste for users comfortable with the terminal.

For the manual method, create the .claude/skills/ directory if it doesn't exist: 'mkdir -p .claude/skills'. Then place your downloaded or pasted skill files inside. The naming convention doesn't affect functionality — Claude Code reads all .md files in the skills directory regardless of name. However, using descriptive names (research.md, email-writing.md, qualification.md) helps you stay organized and helps Claude identify which skill covers which topic when there are multiple files loaded.

Installing Claude Skills in Cursor

Cursor is a popular AI-powered code editor that uses Claude models and supports a similar skills/rules system. In Cursor, skill-like instructions are called 'Rules' and are stored in the .cursorrules file at the root of your project or in the .cursor/rules/ directory for multi-file rules. To use Claude skill files in Cursor, you adapt them for Cursor's format rather than using the .claude/skills/ directory structure.

For simple installations, you can paste the contents of a Claude skill file into Cursor's .cursorrules file directly. Cursor reads this file and applies the instructions to Claude's behavior in that project. If you want to use multiple skill files in Cursor, the .cursor/rules/ directory approach is better — you can place individual .md files there and Cursor will load them all, similar to how Claude Code handles the .claude/skills/ directory.

Some Claude skill collections include Cursor-specific versions of their skill files, or instructions in their README for adapting the files for Cursor use. Look for a 'Cursor' section in the collection's documentation before adapting the files manually. The adaptation process is usually minimal — mostly removing Claude Code-specific instructions and ensuring the file format matches Cursor's expectations.

Cursor's Rules feature has evolved significantly in recent versions. Cursor now supports project-level rules, global rules, and directory-specific rules, giving you fine-grained control over when each skill file applies. For example, you could have a copywriting skill that only applies in your marketing content directories and a code review skill that only applies in your source code directories. This contextual rule application is something Claude Code's flat skills directory doesn't support natively.

Verifying Your Installation Works

After installing a skill collection, run a quick verification test before relying on the skills for real work. Start a fresh Claude Code session in your project directory and ask Claude a question that specifically tests one of the installed skills. For an outbound sales skill collection, try: 'I need to research [specific company name] as a potential prospect. Please use the prospect research skill to analyze them.' If the skill is properly loaded, Claude should follow a structured research process matching the workflow defined in the skill file, rather than giving a generic response.

Another verification approach is to ask Claude what skills it has available: 'What skill files do you currently have loaded in this session?' Claude should be able to list the skill files and describe what each one covers. If Claude reports no skills loaded, or if the list doesn't match what you installed, there's a path or configuration issue to debug.

Compare the output format to the expected output format defined in the skill file. Open the skill file you're testing in a text editor and look for sections describing output format, templates, or examples. Then run a test task and check whether Claude's output matches that structure. If the output is generic and doesn't follow the defined format, the skill file may not be loading correctly, or the skill's instructions may need refinement.

For ongoing verification, periodically run a standard test task — like researching the same company or writing an email for the same fictional prospect — and compare outputs across sessions. Consistent output format and quality indicates the skills are loading reliably. Inconsistent output may indicate a configuration issue, a context window saturation problem, or skill files that need to be made more specific.

Troubleshooting Common Issues

The most common installation issue is skills not being found by Claude Code. This usually means the skill files aren't in the right directory. Claude Code looks for skill files in .claude/skills/ relative to the current working directory when you start Claude Code — so if you're running Claude Code from a different directory than where you cloned the skills, Claude won't find them. Always start Claude Code from the project directory that contains the .claude/ folder, or use absolute paths to verify where your skills are installed.

If skills are loading but Claude isn't following them, the most common cause is context window saturation — especially if you have many or very large skill files. Try removing some skill files temporarily and running the same test again. If Claude follows the skills more precisely with fewer files loaded, you've confirmed the saturation issue. The solution is to prune your skill collection to the most essential files for your current work session, or to create condensed versions of large skill files that capture the key frameworks in fewer words.

Skill files with formatting errors or encoding issues can fail to load properly. Open the problematic file in a text editor and look for unusual characters, broken Markdown syntax, or encoding issues (sometimes introduced when copy-pasting from PDF or Word documents). Plain text Markdown files with standard UTF-8 encoding work most reliably. If you're unsure, create a new file and manually type or paste the content to ensure clean formatting.

Conflicting instructions across multiple skill files cause Claude to produce inconsistent output. If you notice that Claude seems to 'choose' different frameworks on different runs, open your skill files and look for contradictory instructions about the same topic. For example, if one skill file says 'emails should be under 100 words' and another says 'include at least three value proposition sentences,' Claude will receive conflicting signals. Resolve conflicts by editing one or both files to align on a single standard.

Next Steps

Once you have your first skill collection installed and verified, take a few minutes to review each skill file and customize it for your specific context. Find the persona definition sections and update the target company size, industry, buyer role, and pain points to match your actual ICP. Update any value proposition or positioning language to match your product's actual differentiators. These small customizations dramatically improve the relevance and quality of Claude's output.

Consider setting up a personal skill library in a private GitHub repository. As you refine skill files and potentially write new ones, having a private repo lets you version your customizations and share them across multiple projects or machines. Name the repo something like 'my-claude-skills' or 'my-gtm-skills' and push your customized skill files there. Then when you start a new project or switch machines, you can clone your personal skill library in minutes and pick up right where you left off.

Explore the GTM Skills Directory to discover other skill collections that complement what you've installed. Most GTM practitioners eventually settle on a combination of 2-4 skill collections covering different aspects of their work: one for research, one for writing, one for strategy, and one for operations. Building this stack gradually — adding one collection at a time, integrating it with what you already have, and verifying it works well together — is more effective than installing everything at once.

Finally, consider contributing back to the open-source skill community. If you write a skill that works particularly well for your use case, or if you significantly improve an existing skill, consider publishing your work. You can fork the original collection on GitHub, commit your improvements, and submit a pull request to the original author. If your improvements are specific to your company context, publish them as a separate collection with clear documentation. The GTM skill ecosystem grows through exactly this kind of contribution from practitioners who are actually using the tools in the field.

If your primary use case is outbound prospecting or B2B sales, Vibe Prospecting's Claude skill collection is one of the best starting points available. It's actively maintained, designed specifically for Claude Code, and comes with a companion data integration that gives Claude access to live business profiles — turning the installation from a one-time setup into a genuine competitive advantage for your outbound motion. Find it at the GTM Skills Directory and start with the ICP research and email personalization skills.