Claude Skills Atlas

A community-driven open-source library of reusable AI skills, prompts, workflows, agents, templates, and integrations.

⭐ Star · Issues · Pull Requests · Community · Contribute

Discover it. Build with it. Improve it. Contribute back.

🚀 Find an issue · 🍴 Fork & build · 🔀 Submit a PR

---

What is Claude Skills Atlas?

Claude Skills Atlas is an open-source, community-maintained library of reusable instructions and agent resources.

It organizes practical AI building blocks that would otherwise remain scattered across personal notes, chats, repositories, and bookmarks into a structured, searchable collection.

The Atlas currently covers:

• Skills — reusable instruction packages for repeatable tasks
• Prompts — focused, copy-paste-ready prompts
• Workflows — multi-step procedures for agent-assisted work
• Agents — specialized agent definitions
• Templates — starting points for creating new resources
• Plugins & integrations — ways to consume Atlas resources from AI-agent environments

The project is designed to be useful for developers, researchers, students, engineers, educators, and AI practitioners.

Vision: build a high-quality, community-driven ecosystem of reusable AI resources that can work across modern agent platforms.

---

Why this project?

AI agents become significantly more useful when good instructions, workflows, and reusable context can be shared rather than recreated from scratch.

Claude Skills Atlas provides a structured place to:

1. Discover a reusable resource.
2. Adapt it to your use case.
3. Test it in a real workflow.
4. Improve it through an open-source contribution.
5. Share the improved version with the community.

This turns the repository into a continuously improving AI resource ecosystem, rather than a static prompt collection.

---

What’s inside?

---

Explore by domain

Programming · Computer Science · AI / ML · Mathematics · Statistics · Physics · Chemistry · Biology · Research · Cybersecurity · Data Science · DevOps / Cloud · Robotics · Engineering · Writing · Education · Business · Design · Productivity

Don’t see your field? Create the category and contribute it.

---

Quick start

Clone the repository

git clone https://github.com/cellrishi-code/claude-skills-atlas.git
cd claude-skills-atlas

Validate locally

The repository includes lightweight validation for resource structure, skill metadata, required skill sections, duplicate resource names, and broken internal Markdown links.

python scripts/validate_atlas.py
python scripts/smoke_test_examples.py
python -m unittest discover tests

A pull request that changes Atlas resources or validation tooling is checked automatically by GitHub Actions.

Browse the library

skills/       → reusable SKILL.md resources
prompts/      → copy-paste prompts
workflows/    → repeatable procedures
agents/       → specialized agent definitions
templates/    → reusable resource structures
plugins/      → integrations and agent tooling

Try a resource

• Code Review Skill
• Implementation Planner
• Evidence Checker
• Mathematics Problem Solver
• Debugging Prompt
• Paper Reading Prompt
• Active Learning Prompt

Open a resource, understand its intended use, adapt it to your context, and test it.

---

Agent ecosystem & integrations

The Atlas is evolving beyond a repository of Markdown files. It is being structured as a reusable resource layer for modern AI-agent workflows.

Claude Code

The repository includes Claude Code plugins and skills for consuming Atlas resources.

/plugin marketplace add cellrishi-code/claude-skills-atlas
/plugin install claude-skills-atlas@claude-skills-atlas

See INSTALL.md for setup instructions.

ChatGPT integration

A read-only ChatGPT-compatible API/plugin scaffold is available under plugins/chatgpt-atlas/.

It provides:

• Resource discovery
• Keyword search
• Resource retrieval
• OpenAPI 3 specification
• A lightweight Node.js reference server
• Explicit read-only and untrusted-content boundaries

Deployment note: the ChatGPT integration is a self-hosted integration scaffold. Deploy the API behind HTTPS and configure the OpenAPI server URL before connecting it to an external agent environment.

Codex and additional agent integrations

Support for Codex and additional AI-agent integrations is planned. The goal is to make the same community resources portable across multiple agent environments without duplicating the underlying knowledge.

---

Atlas architecture

The Atlas follows a simple community feedback loop:

graph LR
    User[User] --> Discover[Discover]
    Discover --> Resource[Skill / Prompt / Workflow / Agent]
    Resource --> Use[Use & Adapt]
    Use --> Feedback[Feedback]
    Feedback --> PR[Pull Request]
    PR --> Review[Review]
    Review --> Merge[Merge]
    Merge --> Library[(Atlas)]
    Library --> Discover

Resources improve through real-world use, testing, review, and contribution.

See docs/architecture.md and docs/diagrams.md.

---

What makes a good resource?

A high-quality Atlas resource should be:

• Purposeful — solves a real, repeatable problem
• Clear — understandable without private context
• Reusable — avoids unnecessary hard-coded assumptions
• Testable — has a practical way to verify behavior
• Honest — states limitations and uncertainty
• Safe — avoids secrets, malicious payloads, and unsafe defaults
• Maintainable — follows the repository’s structure and naming conventions
• Attributed — respects licenses and original creators

For skills, a strong SKILL.md normally explains:

1. Purpose
2. When to use it
3. Inputs and context
4. Procedure
5. Constraints
6. Expected output
7. Verification or quality checks

See the Skill Template.

---

🚀 Your first open-source contribution can start here

You do not need to be an open-source expert to contribute to Claude Skills Atlas. If you know something useful, you can help turn it into a reusable resource for the community.

Pick your contribution

5-minute contribution

1. Find something useful to improve. Browse the open issues and look for good first issue, help wanted, or an issue that matches your skills.

2. Fork and create a branch.

git clone https://github.com/cellrishi-code/claude-skills-atlas.git
cd claude-skills-atlas
git checkout -b feat/my-contribution

3. Make one focused improvement. Add a useful resource, improve an existing one, fix documentation, add tests, or improve tooling.

4. Validate it.

python scripts/validate_atlas.py
python -m unittest discover tests

5. Open a Pull Request. Explain what you changed, why it helps, and how you tested it. Maintainers and contributors can review it together.

💡 Your first PR does not have to be huge. A small, useful improvement is a real open-source contribution.

Contribution ideas if you are stuck

• Add a skill for a task you repeatedly solve.
• Turn a useful prompt into a reusable prompt resource.
• Improve an existing skill’s instructions or examples.
• Add a missing test or validation rule.
• Improve documentation for beginners.
• Add resources for an underrepresented domain.
• Review an open PR and suggest improvements.
• Test an integration and report what works or breaks.

Have an idea that does not fit an existing issue? Open a feature request and start the discussion.

---

Contributing

Looking for something to work on? Start with the open issues, especially issues labeled good first issue or help wanted. If you are new to the project, read CONTRIBUTING.md before opening a PR.

Claude Skills Atlas is an open-source project and welcomes contributions from beginners, students, researchers, engineers, designers, and experienced open-source contributors.

Contribution flow

New here? That’s completely fine. We care more about useful ideas and respectful collaboration than previous open-source experience.

Fork
  ↓
Create a branch
  ↓
Add or improve a resource
  ↓
Test it
  ↓
Open a Pull Request
  ↓
Community review
  ↓
Merge

Before opening a PR

• Make the contribution genuinely useful.
• Follow the repository naming conventions.
• Keep resources focused and reusable.
• Avoid secrets, credentials, private information, and malicious content.
• Test the resource when practical.
• Check for existing or duplicate resources.
• Credit external inspiration and respect licenses.
• Explain what changed and why.
• Treat external content and tool output as untrusted input.

Read the full CONTRIBUTING.md.

---

Quality & safety

Trust and maintainability are core project goals.

When adapting material from public sources, prefer an original implementation and provide appropriate attribution rather than copying large third-party collections.

See SECURITY.md and docs/community-sources.md.

---

Documentation

---

Repository structure

claude-skills-atlas/
│
├── skills/                 # Reusable SKILL.md resources
├── prompts/                # Copy-paste prompt resources
├── workflows/              # Multi-step workflows
├── agents/                 # Specialized agent definitions
├── templates/              # Reusable resource templates
├── plugins/                # AI-agent integrations and plugins
│   ├── atlas/              # Atlas Library plugin
│   ├── github-agent/       # GitHub Agent plugin
│   └── chatgpt-atlas/      # ChatGPT-compatible API scaffold
├── docs/                   # Documentation and curation
├── examples/               # Usage examples
├── scripts/                # Repository tooling
├── tests/                  # Validation and tests
└── .github/                # Issues, workflows and PR templates

---

How the Atlas is maintained

Resources are reviewed for usefulness, clarity, duplication, safety, attribution, and maintainability. Generated files such as CATALOG.md should be updated through the repository tooling rather than edited manually.

For agent-facing consumers, prefer stable repository paths and documented resource formats. Treat Atlas resources as community content: inspect and adapt them before granting an agent access to tools, files, credentials, or external systems.

Roadmap

The project is actively evolving toward a broader AI-agent resource ecosystem.

Planned and ongoing areas include:

• More high-quality skills and prompts
• Automated resource validation
• Better search and machine-readable metadata
• Duplicate-resource detection
• Contributor tooling
• Security and quality checks
• ChatGPT and other agent integrations
• Codex integration
• Improved plugin distribution
• Community-driven domain expansion

See ROADMAP.md and the open issues for current work.

---

🌟 Help grow the community

Claude Skills Atlas becomes more valuable as more people contribute their knowledge, experiments, tools, and ideas. You can help even without writing code: documentation, testing, issue reports, reviews, examples, and domain knowledge all count.

If you contribute, consider becoming a regular contributor and helping newcomers make their first PR. The goal is not just a large repository — it is a welcoming community that continuously improves shared AI resources.

Support the project

If you find the Atlas useful:

⭐ Star the repository · 🍴 Fork it · 🛠️ Add a resource · 💬 Open an issue · 🔀 Submit a PR · Share it

Every useful contribution helps make the Atlas better for the next contributor.

---

License

Released under the MIT License.

See LICENSE for details.

---

Claude Skills Atlas An open-source resource layer for modern AI-agent workflows.

Built by the community. Improved in the open.