Repository directory
What each project does
Search by name or filter by role. Every card includes the real use case, expected impact and the person it best serves.
Repomix
yamadashy/repomix
Packages an entire local or remote repository into one AI-friendly document so Claude can reason about the project as a whole instead of seeing isolated files.
How it worksScans a repo, respects ignore rules, estimates tokens and exports XML, Markdown, JSON or plain text.
Why it mattersExcellent for architecture reviews, migrations, onboarding, refactors and “understand this codebase” prompts.
Watch outReview the output for secrets and avoid sending sensitive proprietary code to an unapproved model.
Everything Claude Code
affaan-m/ECC
A broad Claude Code configuration framework containing specialized agents, skills, commands, hooks and opinionated workflow patterns.
How it worksInstalled as a Claude Code plugin or copied selectively into your own project and user-level configuration.
Why it mattersTurns a blank Claude Code setup into a structured operating environment for planning, testing, reviewing and shipping software.
Watch outDo not install everything blindly; adopt only the rules and automations that match your stack.
Dify
langgenius/dify
A production-oriented platform for building LLM applications, agents and RAG workflows through a visual canvas, APIs and an operations layer.
How it worksConnects to many model providers, knowledge bases, tools and workflow nodes, then exposes finished apps through a UI or API.
Why it mattersUseful when a prototype must become a managed product with prompts, datasets, logs, evaluation and deployment controls.
Watch outIt is a full platform, not a small Claude add-on; self-hosting introduces infrastructure and security responsibilities.
Flowise
FlowiseAI/Flowise
A visual, node-based environment for assembling AI agents, RAG pipelines and multi-step flows with less custom glue code.
How it worksDrag nodes onto a canvas, connect models, prompts, memory, vector stores, tools and APIs, then expose the flow as an endpoint.
Why it mattersGreat for quickly validating an agent architecture before committing to a fully coded implementation.
Watch outVisual simplicity can hide operational complexity; test authentication, tool permissions and input validation carefully.
Onyx
onyx-dot-app/onyx
A self-hostable AI workspace for chat, enterprise search, RAG, agents and actions connected to internal company knowledge.
How it worksIndexes connected sources such as workplace documents and apps while preserving a centralized conversational interface.
Why it mattersDesigned for organizations that want an internal AI assistant grounded in their own information and deployable in controlled environments.
Watch outPermission syncing, connector scope and data retention need enterprise-grade governance.
Claude Skills by Anthropic
anthropics/skills
Anthropic’s public collection of Agent Skills: folders containing instructions, scripts and resources that Claude loads for specialized, repeatable tasks.
How it worksEach skill describes when it should trigger and gives Claude a structured workflow, supporting material and optional executable helpers.
Why it mattersProvides both practical examples and the clearest reference architecture for building your own reliable skills.
Watch outSkills can contain scripts; inspect every file and understand what permissions the agent will receive.
Awesome Claude Skills
ComposioHQ/awesome-claude-skills
A curated directory of community-created Claude skills and examples across development, content, research, automation and business workflows.
How it worksActs as a discovery index rather than one monolithic plugin; you browse the list and install or adapt individual skills.
Why it mattersSaves time when exploring what the community has already built and reveals useful patterns for your own skill library.
Watch out“Curated” does not mean audited. Check maintainers, code, license, recent activity and requested permissions.
Obsidian Skills
kepano/obsidian-skills
Agent skills that teach Claude how to work effectively with Obsidian vaults, including Markdown, Bases, JSON Canvas and command-line workflows.
How it worksProvides conventions and task procedures so an agent can create, reorganize and connect notes without destroying vault structure.
Why it mattersBridges a personal knowledge system and an AI agent, enabling repeatable research, synthesis and content workflows.
Watch outUse version control or backups before allowing bulk edits across a valuable vault.
NotebookLM Skill
PleasePrompto/notebooklm-skill
Connects Claude Code with Google NotebookLM so the agent can query notebooks and return answers grounded in the documents loaded there.
How it worksUses a browser-assisted integration and persistent session state to interact with NotebookLM libraries and notebooks.
Why it mattersCombines Claude’s workflow automation with NotebookLM’s source-grounded research behavior and citations.
Watch outBrowser automation and stored authentication are sensitive; isolate credentials and review the implementation before use.
Marketing Skills
coreyhaines31/marketingskills
A focused collection of marketing skills for Claude Code and other compatible agents, covering CRO, copywriting, SEO, analytics and growth engineering.
How it worksEncodes repeatable marketing frameworks and task-specific instructions into installable skill folders.
Why it mattersReduces generic output by giving the agent a structured process for audits, campaigns, positioning and conversion work.
Watch outFrameworks improve consistency but do not replace first-party customer data, legal review or human brand judgment.