Skills: The Missing Link Between Agents and Organizational Memory

Let’s talk about skills.

Skills are all the rage right now for agents. A skill is a specialized task you teach an agent so that, suddenly, it’s smarter than before.

At its core, a skill is very simple: a small piece of text or markup that shows the agent how to do a specific activity in a bespoke way. For example, if your agent doesn’t know how to write entries into a PDF form, you can teach it how to write to a PDF file.

Skills are powerful because everyone has their own workflow and unique way of getting work done.

The Real Challenge: Skills at Scale

In the real world, deploying skills at scale is easier said than done.

In large enterprises, there is a vast amount of bespoke, knowledge-based logic trapped inside code. Imagine a repo with millions of lines of code and 10 years of history. Do you really want your team to rediscover the same techniques for getting work done over and over?

If you hand that repo to an agent with no prior history, it will falter at best. It will repeat the same setup steps for everyone in the company, on the same machines, to get the same things done again and again.

That has massive cost implications. Every time an agent does something bespoke, it doesn’t automatically create memory. The next teammate has to rediscover the same path.

In some ways, skills are organizational memory.

Growing a Living Skill Set

Now imagine a brownfield project with tens, if not hundreds, of engineers working on a repo with a long history.

What if I told you that you can not only get skills created for you, but—more importantly—as time goes on and the repository requires new skills, your basket of skills can grow over time?

If you’re curious to learn more about this, come talk to us at Exceeds.