Bluebag AI

When Anthropic published their Skills system, it clicked for me instantly: Give agents a filesystem-based “skill library” of instructions, scripts, and reference files, and let it progressively load what it needs. Sadly, in my own projects, I wasn’t using Claude (most workloads were on Gemini, mostly for cost and flexibility). So I couldn’t use Anthropic’s Skills directly, but I really wanted that architecture in my remote agents built on Vercel's ai-sdk. So I built an Anthropic-style Skills infrastructure that runs with any remote agent. It handles all the difficult parts of running skills remotely at scale - sandbox management, skills reusability, installing dependencies, concurrent file downloads and a lot more Right now, it lets you: - Bundle metadata, instructions, reference files, and scripts into a Skill directory -Run Python or JS scripts inside Skills (with automatic package installation) -Use a files API so the model can create files, reference them, mint temporary download links, and so on - Manage everything via a CLI (push/pull), a TypeScript SDK, and a small web app for API keys, PATs, and a playground Here's a link to the playground: https://www.bluebag.ai/playground Curious to see what you build on the generous free tier!