Claude Code Skills for VOSS/FE, EXOS/SE, Switch engine and GNS3

[Salva_ferrer]

Hi all,

This is not a question or request for help, I am here to write about my exciting trip using Claude Code to manage EXOS and Site Engine. In the not too far future, Control and Analytics.

The usual disclaimer applies, keep an eye on what Claude is doing, approve everything, my advise for any complex task ask code to build a plan, review it and only then launch the execution.

I started with a concept to use CloudIQ API to build some guest management app using ppsk, then I thought, what if Claude helps me build a GNS3 lab for a customer training. Last week I finished my first complete switch deployment co-authored with Claude Code, I provided description of what I needed and Claude implemented, verified and diagnosed when something was not as expected.

All in all deployment has been flawless, and I built the best documentation to close a project in my 25 years career, because Claude doesn't miss a CLI keystroke nor forgets why you set up something the way you did.

My next challenge, if somebody wants to chime in is, can claude code pass an ECA lab? The answer to this in a few months.

Fruit of the story above is the document and skills attached. It describes how to use EXOS/VOSS/SE in a gns3 environment, from importing the templates. to building test nodes attached to the network to test endpoints. It has saved me countless hours diagmosing and testing concepts, like an advanced ToIP/WIFI-AP UPM script in python, with safeguards for deployment specific issues I found. This experience also showed how quick Claude Code can build and diagnose UPM using python or TCL, which can allow to move SE and FE platforms to a next level as fully automated programmable network platforms.

I'm sharing in the hope somebody find it useful in his job or sparks some investigation to improve our ability to build a fully programmable and replicable, deployment and  diagnosys model. We don't need EP1  for it.

since I cannot drop here the full zip package, here you have the link to my repo and the pdf with the main description.

https://github.com/salva-ferrer/ExtremeNetworks

Have fun!!

Salva

[Salva_ferrer]

Another side project of mine... I have a subtle knowledge of site engine/cloudiq Api 😜 . What can we build in terms of ecosystem integrations if we face Claude with that Api and any tool you need to link to in your customers' environment?

If anybody has a preferred one, Le me know.

Regards

[cochin78mark]

Hello,

Your workflow optimization—treating Claude like a junior engineer who requires a solid architectural plan before hitting execution—is spot on. As you noted, the real superpower here isn't just speed; it's consistency. Eliminating human typos in CLI commands, retaining context across a massive configuration lifecycle, and building airtight documentation natively are massive wins for project handovers and post-mortems (especially after 25 years in the trenches!).

Best Regards

[Brent_Addis]

LOL did you just reply to a hand written post about AI... with AI? That's amazing.

[Salva_ferrer]

Indeed my experience... Ai has been trained with any python example you can imagine so writing a piece of python code... Piece of cake.

But general networking knowledge is laking in its training corpus, I found myself arguing basic stp concepts that a junior grasps after diagnosing a couple of stp recalculations but Claude had to be stopped in its planning and asked to read 802.1d before continuing, literally.

The thing is... You can ask it to do it and after that it has all the required knowledge for the remaining of the project.

And definitely, documentation... If you are verbose explaining your decisions, nothing gets lost, mix it with a good note taking app and it's a killer combination. 

I must also say I've been lucky because I'm a Linux Desktop user, integrating minicom, ssh password storage has been flawless.

OH, one last thing i forgot to mention. Get a good knowledge management app (there are many integrated with Claude) that way whatever it learns in one project is also available manually or automatically to other projects.