# borghei/Claude-Skills: 368 skills, 76 agents and 859 Python tools for AI assistants

> A large, multi-domain skills library that installs into Claude Code, Cursor, Codex, Gemini CLI and Copilot. It is strongest on regulated industries and project management, and its install path is a single npx command.

**borghei/Claude-Skills** — 368 AI skills, 76 expert agents, and 859 stdlib Python tools for every team: engineering, PM, marketing, C-level, compliance, business ops, and research. Installs on Claude Code, Cursor, Codex, Gemini, Copilot, and 6 other AI assistants.

- Repository: https://github.com/borghei/Claude-Skills
- Website: https://borghei.github.io/Claude-Skills
- Stars: 846 · Forks: 154
- Language: HTML
- License: NOASSERTION
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/borghei-claude-skills

## What borghei/Claude-Skills is for

Most skill collections for AI coding assistants stop at the codebase. borghei/Claude-Skills starts from the opposite premise: the README says it is "AI skills for every team, not just engineering." The repository ships 368 skills across 20 domains, 76 expert agents, and 859 Python tools, and the domain list is the tell. Engineering has 91 skills, but project management has 68, marketing 39, C-level advisory 31, and compliance 27. There are also smaller domains for legal, HR, sales, finance, research and research operations.

The intended user is not only a developer. The README describes workflows such as shipping a SOC 2 audit, running a sprint retro, drafting a board update and standing up a fintech compliance program. That framing assumes someone who already has an AI assistant open and wants a structured procedure rather than a blank prompt. The compliance coverage is the most distinctive part: the README claims 18 frameworks, naming SOC 2, ISO 13485, MDR, FDA, EU AI Act, NIS2, DORA, HIPAA and GDPR, and calls itself "the only OSS skills library covering regulated industries." That is a claim from the project, not an independently verified one, and it is the kind of statement worth testing against your own framework list before you rely on it.

## How the skills, agents and Python tools fit together

The repository separates three layers. Skills are the unit of work. Agents are personas that pull several skills together. Tools are Python scripts the skills can call.

At the agent layer, the README describes 8 cross-domain personas (startup-cto, growth-marketer, solo-founder, content-strategist, devops-engineer, finance-lead, product-manager, skill-creator) that combine skills from multiple domains. Beneath them sit 68 single-domain cs-* agents, with examples like cs-cto-advisor, cs-fintech-advisor and cs-scrum-master. The README states you activate a persona with /persona startup-cto or by mentioning it in conversation. That is a meaningful design choice: you are not selecting a skill per task, you are selecting a role and letting it choose.

The tool layer is where the project makes a concrete engineering promise. The README says the 859 Python tools are standard library only, with "no pip cliffs," so they can be piped into CI, MCP servers or your own scripts. Standard-library-only is a real constraint that cuts both ways: it means no dependency resolution and no virtualenv, but it also means the tools cannot use third-party libraries for parsing, HTTP clients or data formats.

Above that, the repository defines orchestration patterns. The README lists four multi-agent patterns (sequential pipeline, parallel fan-out, supervisor delegation, consensus voting) defined in a standard under standards/. There are also 26 slash commands for common workflows, including /tdd, /rice, /prd, /retro, /tech-debt, /security-scan, /sprint-plan and /changelog. The README says running /README in Claude Code prints the full list. The repository layout backs this up: there are commands/, pipelines/, standards/, templates/, tools/ and evals/ directories at the top level, plus a registry.json and skills.json.

## Installing borghei/Claude-Skills and running a first skill

There are two install paths, and they suit different people. The no-terminal path is aimed at chat users. The README says to open the skill library site, click "Try it now" on any skill, and paste the prompt into Claude.ai, ChatGPT or Gemini. It also describes a "Save to my AI" button that installs a skill as a Claude Project or Custom GPT.

Inside the Claude desktop app, the README gives a plugin route: open Customize in the bottom-left, go to Browse plugins, and enter the repository path.

```bash
borghei/Claude-Skills
```

All 368 skills then become available across Claude conversations, according to the README.

For developers, the documented install is a single npx command. The README states it auto-detects Claude Code, Cursor, Codex, Gemini CLI, Copilot, Windsurf, Cline, Aider and Goose, and drops the skill into the right directory for your assistant. It requires Node 18+ and states that no Python is required.

```bash
npx @borghei/claude-skills add senior-fullstack
```

Replace senior-fullstack with the skill you want. The README points to docs/INSTALLATION.md for the Claude Code plugin, MCP server, manual install and CLI tooling, and to bundles.json for role-based bundles. A bundle is the faster first move if you do not yet know which individual skills you need. The README lists bundles named SaaS Founder Kit, DevOps Kit, Compliance Kit, Growth Kit, Product Kit, Data Kit, Security Kit, Finance Kit, PM Starter Pack and PM Career Track, all defined in bundles.json.

One caveat on the CLI: the README does not document a rollback or uninstall command, nor does it describe what happens when a skill already exists in the target directory. Verify what the CLI writes before running it in a repository you care about.

## The compliance and project management domains carry the weight

Two domains justify the repository's size more than the others.

Compliance is the differentiator. The README states there are 18 compliance frameworks with checklists, gap analyses and audit-ready outputs, and the domain directory is ra-qm-team/. If your team works against SOC 2, ISO 13485, MDR, FDA, EU AI Act, NIS2, DORA, HIPAA or GDPR, a skill that produces a gap analysis is a different object from a prompt template. The trade-off is that audit-ready output is a strong phrase. The README does not explain how the checklists are maintained against framework revisions, and a compliance framework that drifts out of date is worse than no checklist at all. Treat the outputs as drafts to be reviewed by whoever owns the control, not as finished evidence.

Project management is the deepest domain by the project's own account: 68 skills, each with a worked example, an anti-pattern library and a deterministic scoring rubric, according to the README. Fifteen PM skills are callable as MCP tools, and the README lists 5 chain pipelines and 3 live data adapters for Jira, Linear, Notion, Productboard and Confluence. The presence of evals/ and a deterministic rubric is the most interesting signal in the repository: scoring rubrics are testable in a way that prose prompts are not, and the release history shows the PM domain growing deliberately, with v4.7.0 described as making PM "a production toolkit (54 skills, ~30k lines added)." If you are evaluating this project, the PM domain is where the engineering discipline is most visible.

## Where Claude-Skills is the wrong tool

Scale is the first problem. 368 skills across 20 domains is not a curated set, it is a catalogue. If your goal is a small number of skills you can read in full, reason about and audit, this repository works against you. The README's own answer is bundles, which narrows the install but not the repository you are trusting.

Licensing is the second. The README badge says "MIT + Commons Clause" while the repository metadata reports NOASSERTION, meaning no standard licence identifier was detected. Those are not the same statement, and the Commons Clause is not a standard open source term. If you plan to use these skills commercially, read the LICENSE file directly rather than relying on the badge.

Third, the Python tools are standard library only. That is a deliberate constraint, but it means anything requiring a third-party parser, SDK or client has to be reimplemented or wrapped by you. For simple file and text processing that is fine. For anything that talks to a service with a real SDK, expect to write the integration yourself.

Finally, there is the question of which assistant you are on. The README marks Claude Code, Codex, Gemini CLI, Cursor, GitHub Copilot, Windsurf, Cline and Goose as Native, and Aider, Jules and RooCode as Compatible. Native and compatible are different promises, and the README does not define the difference. If you are on Aider, Jules or RooCode, budget time to check what actually lands in your configuration directory.

## Compared with writing your own skills or using a single-vendor set

The obvious alternative is Anthropic's own skills documentation and whatever skills ship with your assistant. The difference is scope and control. Vendor-provided skills are maintained by the vendor, cover the vendor's own integration surface, and carry no third-party licence question. borghei/Claude-Skills trades that for breadth: 20 domains, 11 platforms, and domains like HR, legal and compliance that no assistant vendor is likely to ship.

The second alternative is writing your own. A skill is a prompt plus supporting files, and a team with strong opinions about its own process will get more from ten skills that encode its actual conventions than from 368 that encode someone else's. The counterargument is the tooling layer. Writing a skill is easy; writing 859 standard-library Python tools, a registry, an eval directory and four orchestration patterns is not. If you want the tools and the orchestration patterns, adopting the repository and pruning is cheaper than rebuilding.

The third alternative is a domain-specific tool. If your only need is SOC 2 evidence collection, a dedicated compliance platform will track framework revisions and produce auditor-accepted artefacts in a way that a markdown checklist in a Git repository will not. This repository is strongest when you need many shallow-to-medium workflows in one place, not one deep workflow done properly.

## Maintenance, upgrade cost and licence questions

The repository is not archived, and the last push was on 2026-08-12. The release cadence visible in the release notes is fast: v4.2.0 on 2026-05-04, v4.7.0 on 2026-05-22, and v4.8.0 on 2026-05-27, with v4.8.0 adding 43 skills, a new research domain and 5 C-suite advisors. Between the end of May and mid-August the repository received pushes but no listed release, so the picture is a project that ships in bursts.

That cadence is the upgrade cost. A library that added 43 skills in a single release will change the contents of domains you depend on. If you install via npx, you are pulling whatever the current state is. There is no documented pinning mechanism in the README, so if you need reproducibility you should vendor the specific skill directories into your own repository and update them deliberately rather than re-running the installer.

On licensing: the README badge reads "MIT + Commons Clause" and the repository metadata reports NOASSERTION. The Commons Clause historically restricts selling software that includes the licensed code, which is a materially different position from plain MIT. The README does not state the terms, so read LICENSE before you build anything commercial on top. This is a description of what the repository says, not legal advice.

## Conclusion

Adopt it if you want a broad, mostly non-engineering skill library that installs with one npx command and covers compliance, project management and business operations alongside engineering. Skip it if you need a small, auditable set of skills you can read end to end, or if you cannot accept the licence terms as published. Before committing, verify the LICENSE file's exact wording, confirm the CLI drops files into the directory your assistant actually reads, and check whether the skills you plan to use have been touched since the last push on 2026-08-12.

## FAQ

### What do Claude skills in borghei/Claude-Skills actually do?

They package a structured procedure plus supporting files that an AI assistant loads for a task. The README describes them as real workflows you can finish today rather than prompt templates, and several skills call the 859 standard-library Python tools bundled in the repository.

### How do I install borghei/Claude-Skills from GitHub?

The README gives a developer install as npx @borghei/claude-skills add senior-fullstack, which auto-detects your assistant and drops the skill into the right directory. It requires Node 18+ and states no Python is required. In the Claude desktop app, the plugin route is Customize, then Browse plugins, then borghei/Claude-Skills.

### How do I use borghei/Claude-Skills in Cursor or GitHub Copilot?

The README lists Cursor and GitHub Copilot as Native platforms, and the npx installer auto-detects both. The repository also carries .cursor/, .cursorrules and .github/ directories at the top level, and the README points to docs/INSTALLATION.md for per-platform setup.

### How do I use borghei/Claude-Skills in Codex?

OpenAI Codex is listed as a Native platform, and the npx installer auto-detects it. The repository includes a .codex/ directory and an AGENTS.md file at the top level. Per-platform setup details are in docs/INSTALLATION.md according to the README.

### How do I use borghei/Claude-Skills in ChatGPT?

The README describes a no-setup path for any AI chat: browse the skill library, click Try it now on a skill, and paste the prompt into Claude.ai, ChatGPT or Gemini. It also describes a Save to my AI button that installs a skill as a Claude Project or Custom GPT.

## Sources

- [borghei/Claude-Skills on GitHub](https://github.com/borghei/Claude-Skills)
- [Issues](https://github.com/borghei/Claude-Skills/issues)
- [Project website](https://borghei.github.io/Claude-Skills)
- [README](https://github.com/borghei/Claude-Skills/blob/main/README.md)
- [Releases](https://github.com/borghei/Claude-Skills/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/borghei-claude-skills