Noureddine RAMDI / OpenDirectory: A modular skills library for AI agents with targeted installation

Created Sat, 23 May 2026 20:41:14 +0000 Modified Sat, 23 May 2026 20:41:27 +0000
1024 Words
github-stars python cli ai skills prompt-engineering agent

Varnan-Tech/opendirectory

OpenDirectory addresses a common practical problem in AI agent development: managing and installing domain-specific prompt packages, or “skills,” across multiple agents and platforms. Instead of building or embedding skills directly into an AI model pipeline, it offers a modular, local-first repository of skills that users can browse interactively and install on demand for their chosen agent. This separation of concerns can lead to better developer experience and agility when experimenting with different agents.

what openDirectory does and how it works

OpenDirectory is a Python-based repository that organizes a growing library of AI agent “skills” — essentially prompt packages and associated metadata — into a structured catalog. These skills are intended for use with multiple AI agents such as Claude Code, Codex, Gemini CLI, and others. The repo itself doesn’t run AI models; instead, it serves as a curated skill set that can be plugged into AI agents to extend their domain knowledge and capabilities.

The architecture is straightforward: the repo hosts skill packages as directories with prompt files and metadata. The core interaction happens through an npm CLI tool invoked via npx "@opendirectory.dev/skills". This CLI provides a full-screen terminal UI (TUI) where users can browse skill categories, search for specific skills, and install them targeting a specific agent.

Under the hood, the CLI fetches the latest version of the skills, presents the user with an interactive browser, and allows installation of individual skills by name. Skills can be installed for different AI agents by specifying a --target flag, ensuring the right skill format and prompts are applied for the chosen agent.

Supporting agents include well-known AI models and platforms such as Claude Code, OpenCode, Codex, Gemini CLI, Anti-Gravity, OpenClaw, and Hermes. This multi-agent targeting feature is central to the repo’s value proposition, as it abstracts the differences in prompt expectations and input formats.

the npm CLI and targeted skill installation: a practical approach

One of OpenDirectory’s distinguishing aspects is its hybrid Python-to-Node.js workflow. Although the skill packages themselves are organized in Python, the user-facing CLI is distributed as an npm package. This means users need Node.js installed to run the CLI commands via npx, which fetches and runs the latest CLI version without a global install.

This design choice prioritizes ease of use and up-to-date functionality over a pure Python environment, which makes sense given the CLI’s rich interactive features.

The CLI supports commands such as:

npx "@opendirectory.dev/skills"

which launches the interactive TUI browser. Alternatively, users can list all available skills with:

npx "@opendirectory.dev/skills" list

To install a skill for a specific agent, the command pattern is:

npx "@opendirectory.dev/skills" install <skill-name> --target <your-agent>

For example, to install the show-hn-writer skill for the Claude agent:

npx "@opendirectory.dev/skills" install show-hn-writer --target claude

This explicit targeting helps ensure that the right prompt formats and logic are installed, which is critical when working with different AI platforms that may have unique requirements.

explore the project: structure and documentation

If you want to dive deeper into the repo, the README provides clear guidance on installation options beyond the npm CLI, including manual skill downloads and integration with the Claude desktop app or Claude Code native environment.

The repo itself is organized into skill folders, each containing a SKILL.md file that documents the skill’s purpose, usage, and metadata. This structure helps users and developers understand what each skill does and how to integrate it into their workflows.

While there is no runtime AI server or complex backend, the repo’s value lies in its curated and organized skill catalog combined with an easy-to-use CLI for discovery and installation.

verdict

OpenDirectory is a practical, no-frills solution for managing AI agent skills across multiple platforms. Its strength lies in modularity and the convenience of an interactive CLI that supports multi-agent targeting. The tradeoff is a dependency on Node.js tooling for the CLI, which might be inconvenient for some Python-native environments.

The repo does not provide AI models or prompt execution engines; it’s purely a skill catalog and installer. This means users still need to have their AI agents set up separately. The quality and usefulness of the skills depend on their source and maintenance, as the repo curates but does not vet every skill.

For developers experimenting with AI agents like Claude Code or Codex who want to quickly browse and install domain-specific prompt packages, OpenDirectory offers a tidy and extensible ecosystem. It’s especially valuable if you work across multiple agents and want a consistent way to manage skills.

If you prefer a purely Python environment or need runtime AI model hosting, this repo isn’t a complete solution on its own. But for prompt engineers and AI integrators focused on skill management and agent specialization, it’s a useful tool worth exploring.

# Quick start example
npx "@opendirectory.dev/skills"

# List all available skills
npx "@opendirectory.dev/skills" list

# Install a skill for Claude
npx "@opendirectory.dev/skills" install show-hn-writer --target claude

This concise CLI-driven approach minimizes setup friction and keeps skills up to date without global installs or heavy dependencies. Overall, OpenDirectory is a solid choice if your work involves managing diverse AI prompt skills across different agents and you want a straightforward, user-friendly interface to do so.

→ GitHub Repo: Varnan-Tech/opendirectory ⭐ 325 · Python