Hey, I'm Kyzgor. First, I'd like to just point out how much I love the honeypot, it's clever and, clearly, catches out those trying to push through as many PRs as possible for the bounty. I'm not interested in that race to the bottom, AI slop heavy side of things - I'd like to work on something challenging, some genuine issues that your codebase might have, for free, to test my pipeline end to end.

You can have a look at some of my recent PRs to Permit.io OPAL repo - not yet merged, but in review and you'll see the standard and methodology in action before you decide if it's worth your time pointing me at a specific issue:

&

Anyway, love the ethos and clear quality bar you all set and adhere to, exactly the same mindset I have. Please point out any bugs you have that you'd like fixed (worst case scenario, you just close my PR anyway), or if it's easier, I'm happy to dig through your open issues and propose one for you to approve/assign me to - the harder the better, I enjoy the ones that seem impossible or incredibly difficult to fix.

Sorry for the long message, but wanted to get these points across rather than a generic "I want to work on X issue" and be done with it.

Hi team, do we have MCP sampling on the roadmap anywhere? seems like the one MCP piece that's not implemented yet

feels like a natural fit since the LLM proxy's already there. those completions could run with cost tracking + guardrails, rather than a server hitting a model directly with no oversight.

curious if this is something you'd find valuable.. glad to pick it up if there's interest