What's new in Tessl: global installs, watch mode, GitHub badges, and a unified score

Four new features just dropped. Dru Knox, our head of Product, describes what they do and why they matter in this video. I’ve included a description of these with links to more information below.

Install skills globally

Not every skill belongs in a single repo. You’ll often have context preferences that you carry from project to project. These don't fit neatly into a tessl.json that gets committed alongside your code.

You can now install skills globally with a single flag:

tessl install your-workspace/your-skill --global

Global skills are stored in ~/.tessl/ and are available across all your projects without touching any per-project manifest. When you want to remove one, the same flag applies:

tessl uninstall your-workspace/your-skill --global

More info CLI command reference

Watch local changes automatically

If you've ever developed a skill that's installed across multiple agents, you'll know the loop: make a change, copy the updated files into every agent's folder, restart each session, and check whether the change did what you wanted. Repeat.

The new --watch-local flag eliminates the copy-and-restart step. When you install a local skill with it, Tessl monitors the source directory and automatically syncs any changes across all your agents as you save:

tessl install ./my-skill --watch-local

Leave it running in a terminal while you edit. Your agents pick up the changes without any intervention. The feedback loop that used to take minutes now takes seconds.

Show your eval score on GitHub

If you've put the work into building a skill that genuinely improves agent behavior, written clear triggers, tested it against real scenarios, tuned the instructions, you now have a way to show that on your GitHub repository.

Tessl will generate a badge you can embed in your README that displays your skill's eval score. For anyone browsing your repo and deciding whether your skill is worth installing, it's a quick, credible signal: this wasn't just written, it was tested.

It's the same quality and impact scores that appear in tessl search results, now surfaced directly where your skill lives.

One score to represent skill quality

Tessl has always run two types of evaluations on skills:

• Reviews check your skill against established best practices, structure, clarity, trigger quality, actionability of instructions.
• Task evals run agents through simulated real-world scenarios, comparing performance with and without your skill installed, to measure whether it actually changes behavior.

These two signals answer different questions. Reviews tell you about the quality of your skill. Is well-constructed? Task evals tell you about the impact the skill can make to an agent. Does it work in practice?. Until now, you'd need to look at both scores separately and form your own view.

We've now combined them into a single aggregate score on the registry. One number that blends construction quality with real-world impact. This is visible on every skill's registry page and available to your agents when they're deciding whether a skill is the right fit for a given workflow.

These four features ship today. To get the latest CLI:

tessl update

And to install from scratch:

curl -fsSL https://get.tessl.io | sh

As always, questions and feedback welcome in Discord.