Reflection plugin cause loops

[dzianisv]

I have seen this reflection message a few times

{
  "complete": false,
  "severity": "MEDIUM",
  "feedback": "The agent ran the stripewebhook_failure evaluation multiple times but never completed the task. The evaluation consistently failed (agent didn't complete the support task), and while the agent identified issues and created a GitHub issue (#30), it got stuck in a loop of re-deploying and re-running evaluations against k3s without making meaningful progress on fixing the root cause. The user's final message [10] suggested testing locally instead of via k3s deployments to speed up iteration, but the agent's last response was still just re-running the same evaluation. No fix was implemented, no local testing approach was explored, and the evaluation still fails.",
  "missing": [
    "Stripe webhook failure evaluation passing (agent team completing the task)",
    "Root cause fix for why the agent doesn't complete the stripe support task",
    "Local testing approach as suggested by user in message [10]",
    "GitHub iss


Update reflection prompt to prevent loops