sleuth-io/sx

What it solves

sx solves the problem of "asset drift" in AI teams, where custom prompts, skills, and agent configurations are copy-pasted across repositories or machines, leading to inconsistent versions and no single source of truth. It provides a centralized way to manage, version, and distribute these AI assets across various tools and team members without requiring complex infrastructure.

How it works

The system uses a "vault" concept—a collection of AI assets stored in a synced folder (like Dropbox or Google Drive), a Git repository, or a centralized service (Skills.new). It follows a manifest-and-lock pattern:

1. Manifest (sx.toml): Acts as the source of truth, listing all assets and their installation scopes (e.g., specific users, teams, or repositories).
2. Lock File: A per-user resolved file that determines which assets apply to the current user's environment.
3. Distribution: The sx CLI or app installs these assets into the native formats required by various AI clients (such as Claude Code, Cursor, or GitHub Copilot).
4. Cloud Relay: For web-based tools like claude.ai or chatgpt.com, sx can expose the vault as an MCP endpoint via a relay, keeping the actual content local.

Who it’s for

• AI Engineering Teams: Who need to share standardized prompts, rules, and agents across multiple projects.
• Non-Technical Teammates: Who want to share and use AI skills without needing to use Git or a terminal.
• Developers: Who want to maintain a portable set of AI agents and tools that work across different IDEs and AI assistants.

Highlights

• Zero Infrastructure: Can run entirely off a shared synced folder with no server or accounts required.
• Broad Client Support: Compatible with a wide range of tools including Claude Code, Cursor, GitHub Copilot, Gemini, and web-based LLM interfaces.
• Granular Scoping: Ability to target installations to specific organizations, repositories, paths, teams, users, or bots.
• Portable Agents: Define an agent's prompt and dependencies once and deploy it unchanged across different AI tools.
• Observability: Built-in tools for tracking adoption and token usage (sx stats) and auditing changes (sx audit).