Anyone have an issue where the agent simply ignores grilling answers?

This has been a painpoint for me using these skills, and I'm completely turned off by it.

• It claims contradictions exists when there is none. I clarify, it fails to document the clarification, and it happens again.
• It grills me, I give it very specific answers, contradicting any of it's options and "recommendations". It documents what it recommended anyway. Only to later cite it as a decision when I'm confused about why it's asking questions about something I thought was settled.
• It doesn't stay in it's lane, it often over-reaches for answers either already documented, researched or open tickets that is already classified as the grilling decision to make the call, often going on a tangent, and then documenting things that contradicts the next grilling ticket.
• It often does NOT read the documentation, research already done.
• It also often cites ADR or CONTEXT.MD as the reason for the contradiction, but when pushed, it says there is no contradiction.

Congratulations. You have a skill that asks questions just because.

My requests are often hyper-specific and I very much know WTF I'm doing in the software world. The wizard/probes/research portions are great. But this crap where I have to tell it 10 times about how it should work and when I think we have a mutual understand, and it documents the opposite, only for me to later realize it never really documented what I told it to is insane.

literally had to ask it "Are you sure this is documented?" when it claimed the ticket is now closed with the final decision, only for it to come back to me saying "Oops no, I missed updating the document, only the ticket got updated"

Opus 4.8, Opus 5.5. Same same.

/end rant