have you tried vibium's cli + agent skill?

Yes, it's pretty good! I've also written API harnesses for bot-based browser automation so that you can detect fields to fill in, remember where they are for next time you need them, and then if the webpage changes, re-explore and rewrite the tags to remember for the new form fields.

Spoiler: this is to automate ticket submission to my landlord's half-baked web portal, not some kind of nefarious captcha breaking thing.