I know it has to do with the massive f… up that anthropic has done to … basically the whole planet with their degraded opus 4.7 and obviously as such, replit no different in the ecosystem, but… lately the Replit agent has been hitting me hard on my nerves with this recently introduced bug .
“Agent declined to respond to this request due to a violation of our content policy.”
And it goes on and on, spams of this line, countless…
Worst part ? This 4.7 is acting like a terminator rather than a builder, destroyer on steroids.
Anything that has to do with my cloudflare configs / databases and oauth specific stuff 4/5 prompts are being refused to proceed from the begining, or even worse, the agent starts working on changes and suddenly decides to refuse , when prompted again it acts confused not knowing exactly what was stopped or refused, and starts degrading the plans afterwards…
Instead, whenever i see the refusal message in the middle of the prompt i prefer to just stop it completely and redo the prompt from the start, sometimes it works, sometimes it doesnt, this brought all sorts of problems on my projects where its making your life not just a pain… but something that hurts even more and i cant say it here for obvious reasons, really thick.
So my basic question to every kind builder who is reading this… What do you do when you hit these model blockers knowing that the model is refusing to work for false reasons.
I tried other methods of switching the prompts a little bit, using kimi 2.6 or even just ChatGPT and they seemed to work around… sometimes!
Do we have any update on Replit’s end about this situation ? I havent seen discussions talking about it, anywhere yet