What happens when you organize AI agents like a company? You get 15+ departments, 125+ specialized skills, and a completely new way to think about your coding workflow.
cbrock84/headcount is an open-source GitHub repository that transforms Claude Code—Anthropic's command-line AI coding assistant—into a structured "agent organization." Rather than providing a single, generic AI assistant, it offers a virtual company with 15+ departments and 125+ individually installable skills. Each skill functions like a specialized employee with a specific role, ranging from code review to legal compliance checking.
The name "headcount" is deliberate: it frames each skill as a team member you can hire (install) or skip, depending on your needs.
The concept of a single AI assistant handling everything is appealing but often impractical. A lone agent lacks depth in specialized tasks. Headcount solves this by borrowing a proven structure from the real world: organizational design. Instead of one generalist, you get a team of specialists. The engineering department handles code review and refactoring. The marketing department drafts content. The legal department checks license compliance. Each department contains skills that are focused, tested, and independently installable.
Claude Code, released by Anthropic in 2025, has quickly become a staple for developers seeking AI assistance directly in their terminal. Its extensibility through skills and plugins has spawned a vibrant ecosystem of community-built tools. Headcount sits at the intersection of this ecosystem—not just another plugin, but a full framework for organizing AI capabilities at scale.
Claude Code is Anthropic's command-line interface that lets you interact with Claude AI directly in your terminal. You can ask it to write code, explain complex codebases, refactor functions, run tests, and automate repetitive development tasks. It works alongside your existing tools—Git, your editor, your build system—and can execute commands, read files, and make changes to your project.
Skills are modular capabilities that extend Claude Code's functionality. Think of them as specialized instruction sets or plugins. A skill might contain a detailed prompt template, a script, or a set of instructions that tells Claude how to perform a specific task—like reviewing a pull request for security vulnerabilities or generating a project status report from Git history.
An agent organization is a structured collection of AI agents or skills, organized by function and responsibility. Instead of a flat list of capabilities, an agent organization groups skills into departments, each with a clear purpose. This structure makes it easier to navigate, install, and maintain a large number of skills.
The company metaphor is straightforward: headcount treats each skill as an employee with a job title and a department. The engineering department has skills like "code-review" and "refactor." The marketing department has skills like "social-media-post" and "product-announcement." When you install a skill, you're effectively hiring that employee. When you skip it, you're keeping your headcount lean. This metaphor makes the system intuitive—you already know how a company is organized, so you already understand how headcount is organized.
Key Takeaway: Headcount uses a familiar organizational structure to make a complex system of 125+ AI skills navigable and customizable.
The exact department list is maintained in the repository, but based on the project's description and community discussions, you can expect departments that mirror a real software company:
To give you a concrete sense of what these skills do, here are a few examples:
code-review: This skill instructs Claude to analyze a pull request for code quality, potential bugs, security issues, and adherence to best practices. It produces a structured review with severity ratings.social-media-post: This skill generates platform-specific social media content. Give it a product update, and it produces a LinkedIn post, a tweet thread, and a short-form video script.license-compliance: This skill scans your project's dependencies and flags any licenses that conflict with your project's licensing strategy.status-report: This skill analyzes your Git history—commits, branches, merged PRs—and generates a concise status report suitable for a stakeholder update.ci-pipeline: This skill generates a CI/CD pipeline configuration based on your project's language and framework.Skills are organized in a directory structure that mirrors the department hierarchy. Each skill has its own subdirectory containing instructions, scripts, and documentation. The categorization is consistent: each skill has a clear name, a description, and usage instructions. This makes it easy to find what you need and understand what each skill does before installing it.
Key Takeaway: The department structure isn't cosmetic—it's a functional taxonomy that makes 125+ skills navigable without a search engine.
The modular design is headcount's most important architectural decision. In many AI toolkits, you install everything or nothing. That's a problem because different developers need different capabilities. A solo developer building a small web app doesn't need a finance department. A legal professional using Claude Code for contract analysis doesn't need the DevOps skills.
Modular installation solves this. You install only what you need, keeping your environment lean and your Claude Code responses focused. It also reduces the risk of conflicting instructions between skills.
The installation process is straightforward. While the exact commands may evolve, the general pattern involves copying the skill directory into your Claude Code skills folder. For example:
# Clone the repository
git clone https://github.com/cbrock84/headcount.git
# Copy a single skill into your Claude Code skills directory
cp -r headcount/departments/engineering/code-review ~/.claude/skills/
Some skills may have additional dependencies, such as Python packages or API keys, which are documented in each skill's README.
Once you've installed specific skills, your Claude Code environment becomes tailored to your workflow. If you're a frontend developer, you might install skills for code review, accessibility checking, and design system consistency. If you're a solo founder, you might install skills from engineering, marketing, and sales departments.
The key point is that headcount doesn't force a one-size-fits-all configuration. You're the manager. You decide your headcount.
Key Takeaway: Modularity means you can build a lean, specialized AI team without carrying the weight of 125+ skills you'll never use.
Before you can use headcount, you need Claude Code installed. Anthropic provides straightforward installation instructions:
# Install Claude Code via npm
npm install -g @anthropic-ai/claude-code
You'll also need an Anthropic API key. Once Claude Code is working in your terminal, you're ready for headcount.
git clone https://github.com/cbrock84/headcount.git
cd headcount
Take a moment to explore the directory structure. You'll see the departments, each containing its skills. The README at the root contains the full list of departments and skills, along with any global setup instructions.
ls departments/engineering/ — browse the skills in each department.cat departments/engineering/code-review/README.md — understand what the skill does and its requirements.~/.claude/skills/.After installation, run Claude Code in your project directory and test a skill. For instance:
claude
> Use the status-report skill to summarize the last 2 weeks of commits.
If the skill is properly installed, Claude will follow the skill's instructions and produce the expected output.
Key Takeaway: Installation is a matter of copying files. The complexity is in choosing the right skills for your workflow.
A developer on a team of five installs the code-review skill. Now, before merging a pull request, they run Claude Code with the review skill. Claude analyzes the diff, checks for common bugs, flags security concerns, and suggests improvements. The developer gets a second opinion on every PR without waiting for a human reviewer.
A startup founder uses Claude Code for development. After shipping a feature, they use the product-announcement skill. They describe the feature in one sentence. Claude generates a blog post draft, a Twitter thread, and a LinkedIn update—all in the company's tone of voice.
An open-source maintainer installs the license-compliance skill. When adding a new dependency, they run the skill, which checks the dependency's license against their project's license and flags any conflicts. This prevents legal headaches before they start.
A tech lead installs the status-report skill. Every Friday, they run it and get a summary of the week's activity—commits, merged PRs, open issues—formatted for their manager. No more manual log-scrolling.
A DevOps engineer combines skills from multiple departments. They use the ci-pipeline skill to generate a build pipeline, the code-review skill to validate each PR, and the status-report skill to generate deployment summaries. The result is a semi-automated workflow that covers the entire development lifecycle.
Key Takeaway: Headcount's value grows when you combine skills across departments to automate entire workflows, not just individual tasks.
The sheer number of skills can be overwhelming. You'll need to spend time exploring what's available and reading documentation. The company metaphor helps, but there's still a learning curve.
Each skill you install adds instructions and context that Claude Code must process. Installing too many skills could slow down responses or dilute Claude's focus. The modular design mitigates this, but it's worth being selective.
As an open-source project, headcount depends on community contributions. The quality of skills may vary, and some skills may become outdated as Claude Code evolves. Check the repository for activity before relying heavily on a specific skill.
Key Takeaway: Headcount is powerful but requires curation. Be selective about what you install and stay aware of maintenance status.
Headcount is part of a broader movement toward structured AI agent systems. Instead of a single assistant, developers are building teams of specialized agents. This mirrors how human organizations work: you don't ask one person to do everything; you build a team with complementary skills.
Other extensions exist for Claude Code, but most focus on a single capability or a small set of related features. Headcount's differentiator is its breadth and structure. It's not a single tool; it's a framework for organizing many tools.
As Claude Code evolves, headcount will likely evolve with it. We can expect new departments, new skills, and deeper integration with Claude Code's features. The project may also inspire similar efforts for other AI assistants.
Key Takeaway: Headcount is an early example of a trend that will likely become standard: organizing AI agents into functional teams.
The repository welcomes contributions. To add a new skill, follow the existing structure: create a new directory under the relevant department, write the skill's instructions, and document it. Submit a pull request with a clear description of what the skill does.
Found a bug in a skill? Have an idea for a new department? Open an issue on GitHub. The maintainer and community are active and responsive.
Follow the contribution guidelines in the repository. Write clear documentation for your skills. Test your skills before submitting. Be respectful in discussions.
Key Takeaway: Headcount is a community project. Your contributions make it better for everyone.
cbrock84/headcount is a structured, modular system for extending Claude Code. It organizes 125+ skills into 15+ departments, each independently installable. The company metaphor makes it intuitive, the modular design makes it practical, and the open-source nature makes it continuously improving.
Headcount represents a shift in how we think about AI assistance. Instead of a single general-purpose assistant, we're moving toward teams of specialized agents—each focused, each excellent at its job, and each available on demand. Whether you're a solo developer or part of a large team, headcount offers a way to build the AI workforce you need.
It's an open-source GitHub repository that provides a structured collection of 125+ skills for Claude Code, organized into 15+ departments, each independently installable.
Clone the repository, browse the departments, and copy the skill directories you want into your Claude Code skills folder (typically ~/.claude/skills/).
Yes. Each skill is independently installable. You can install one skill or fifty, depending on your needs.
Skills cover the full software development lifecycle: code review, refactoring, CI/CD, marketing content, legal compliance, project management, documentation, security, and more.
Yes, it's open-source. You can use it, modify it, and contribute to it without cost.
Headcount is designed specifically for Claude Code. Skills may be adaptable to other assistants, but there's no official support.
Updates depend on community contributions. Check the repository's commit history and release notes for the latest activity.
Yes. The project welcomes new skills, bug fixes, and documentation improvements via pull requests.
You need Claude Code installed and an Anthropic API key. Skills may have additional dependencies, documented in each skill's README.
Yes. Each skill directory contains a README with usage instructions, requirements, and examples.
Ready to supercharge your Claude Code experience? Explore cbrock84/headcount on GitHub and start installing the skills that fit your workflow today!