JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 52
  • Score
    100M100P100Q97913F
  • License MIT

Modular skill tree compiler and AI rules manager for developer workspaces. Integrates dynamically with Cursor, VS Code, Windsurf, Claude Code, and Antigravity.

Package Exports

  • @leosdc/skilltree
  • @leosdc/skilltree/bin/skill-add.cjs

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@leosdc/skilltree) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

๐ŸŒฒ .skilltree

Skill Tree Flow

npm version npm downloads GitHub stars GitHub issues License: MIT

The universal, community-driven skill tree repository for AI Agents.

Read this in other languages: Portuguรชs (pt-BR)

.skilltree is a plug-and-play modular framework designed to live inside your workspace root. It serves as a localized capability branch that allows you to inject specialized nodes, automations, and advanced toolsets into any AI agent ecosystem (such as VSCode Copilot, Antigravity, Claude Code, Windsurf, and custom agentic workflows).

The repository is built entirely on the philosophy of crowdsourced orchestration: a massive, ever-growing tree of atomic skills, compiled dynamically to inject specialized rules into your IDE.


๐Ÿš€ Concept & Architecture

Instead of managing fragmented configurations for different AI tools, you clone .skilltree directly into your active project directory as a hidden configuration module:

git clone https://github.com/Leosdc/skilltree.git .skilltree

Once embedded, the CLI compiler scans your workspace dependencies, compiles active skills into the exact local files your AI agents read (.cursorrules, CLAUDE.md, .windsurfrules, .copilot-instructions.md, etc.), and automatically synchronizes these rules with your global user-level AI configurations (supporting Antigravity IDE, VS Code / Copilot, Cursor, and Windsurf global chats). This guarantees immediate adoption even by autonomous agentic workflows.


๐Ÿ’ก Why .skilltree? (Vs. Traditional Method)

Traditionally, developers manage system instructions for AI agents in a fragmented, manual, and monolithic way. .skilltree introduces a modular, compiled, and team-ready paradigm:

Feature Traditional Method (Manual) .skilltree Way
Agent Fragmentation Developers must manually maintain separate rule files (.cursorrules for Cursor, CLAUDE.md for Claude Code, .windsurfrules for Windsurf, .github/copilot-instructions.md for Copilot). Unified Rule Compilation: The CLI compiler writes the consolidated prompt guidelines into all major rule vectors simultaneously.
Monolithic Rules AI instructions are pasted into one giant, heavy file. This clutters the LLM's context window with unrelated rules (e.g. AWS rules in a frontend task). Atomic & Hot-Swappable Nodes: Habilidades/Rules are organized as atomic directories. Toggle them on and off dynamically via the terminal.
Discovery Passive. Developers must remember to find and copy rule templates. New team members are unaware of project guidelines. Autonomous Scanner: The CLI parses package dependencies (package.json) and project files to automatically recommend skills to activate.
Team Sync Rules are kept locally on each developer's machine, leading to outdated rules and lack of standardization. Git Versioned & Crowdsourced: Lives inside the codebase as a subfolder. Prompt updates are synchronized across the team via Pull Requests.
Security & Compliance Developers paste unverified prompts from the web, risking safety policy bypasses. Audited Tree Catalog: Centralized security policies (e.g., OWASP compliance, secure Docker configurations) are enforced across the team.

๐Ÿ› ๏ธ Unified Cross-Agent Compatibility

.skilltree remains agent-agnostic. It standardizes skill schemas, generating both local workspace vectors and global system integrations:

AI Agent Injection Channel Integration Type
Antigravity IDE (Gemini) Global user_rules.txt & local .cursorrules Dynamically synchronized on startup for autonomous agents to enforce project guidelines.
Cursor .cursor/rules/*.mdc (Scoped) & .cursorrules Supports smart compilation of individual .mdc rule files with glob triggers (frontmatter).
VS Code (Copilot / Chat) Local .copilot-instructions.md & global settings.json Injected into global chat custom instructions and GitHub Copilot workspace settings.
Claude Code / Cowork CLAUDE.md Loaded automatically by Anthropic's Claude Code CLI.
Windsurf Local .windsurfrules & global settings.json Read automatically by the Windsurf AI assistant locally and globally.
OpenClawn / Custom .clawnrules & active-tree.json Loaded by custom local terminal automation tools and orchestrators.

๐ŸŒ Native Internationalization (i18n)

.skilltree supports multiple languages out-of-the-box, allowing teams worldwide to consume prompts and rules in their preferred language:

  • Interactive Selector: When first launched, the CLI prompts you to select your preferred language (e.g., English or Portuguese), which is saved in .skilltree/config.json. You can change it anytime via the CLI menu.
  • Localized Compilation: The compiler automatically reads localized prompt files (prompt.<lang_code>.md like prompt.pt.md) and translations from the skill's metadata (locales in manifest.json), compiling rules into the exact target files in the active language.
  • Isolated Contexts: Keeps prompt definitions separate by language, avoiding duplicate prompts in mixed languages in the LLM's context window.

๐Ÿ“ Repository Structure

.skilltree/
โ”œโ”€โ”€ bin/
โ”‚   โ””โ”€โ”€ skill-add.cjs            # The universal CLI tool (CommonJS)
โ”œโ”€โ”€ tree/                        # Core catalog categorized by stack and scope
โ”‚   โ”œโ”€โ”€ backend/
โ”‚   โ”‚   โ””โ”€โ”€ postgres-expert/
โ”‚   โ”œโ”€โ”€ DevOps/
โ”‚   โ”‚   โ””โ”€โ”€ docker-hardening/
โ”‚   โ”œโ”€โ”€ security/
โ”‚   โ”‚   โ””โ”€โ”€ owasp-audit/
โ”‚   โ””โ”€โ”€ frontend/
โ”‚       โ””โ”€โ”€ react-security/
โ””โ”€โ”€ active-tree.json             # State management containing active skill IDs

๐Ÿ Quick Start

  1. Install the package @leosdc/skilltree in your project:
    npm i @leosdc/skilltree
  2. Open the interactive dashboard:
    npx skilltree

Option 2: Via Git Clone (Local)

  1. Clone the repository as a .skilltree folder inside your project workspace:
    git clone https://github.com/Leosdc/skilltree.git .skilltree
  2. Run the Dashboard using Node.js:
    node .skilltree/bin/skill-add.cjs

๐ŸŽฎ Controls & Commands

Use the keyboard arrow keys (โ†‘/โ†“/โ†/โ†’) to navigate the skills grid and action menu, press [Space] or [Enter] to toggle skill activation, press [d] or [Delete] to permanently delete a custom skill from your catalog (requires confirmation), and select bottom menu actions to run dependency scans or create skills manually.

If you prefer using direct CLI commands instead of the visual interface:

  • Activate a skill: npx skilltree add <skill-id> (or node .skilltree/bin/skill-add.cjs add <skill-id>)
  • Deactivate a skill: npx skilltree remove <skill-id>
  • Delete a skill: npx skilltree delete <skill-id>
  • Run Automated Scan: npx skilltree scan

๐Ÿค Contributing to the Project

We want to expand the Skill Tree into the largest AI orchestration library in the world! If you want to submit new skills, improve security guidelines, or optimize our CLI engine, please check out our Contributing Guide (CONTRIBUTING.md).

We outline our branch naming conventions (feature/, bugfix/), Conventional Commit standards, and prompt quality checklists.


๐Ÿ’– Acknowledgments & Credits

A special thanks to Ric Neves for creating and sharing the amazing library of skills in the flowgrammers-skills repository, which are now integrated into the Skill Tree ecosystem.


For deep documentation on workspace integration, see TUTORIAL.md. For project history, see CHANGELOG.md.