https://github.com/etiennemunnich/crs-agent-skill/ skill I built to help OWASP CRS folk or any web application dev protect their workloads when using modsec / coraza + CoreRuleSets. I wanted to know how I can eval and monitor through the openhands application - I want to know what the skill is doing, how many calls to MCPs/ skill files used, etc etc. What kind of visiblity tooling can I use with openhands?

Hmm, if you use conversation history will give you a broad overview of a run, but then combine with observability tracing via OpenTelemetry (OTEL) which openhands supports and view the traces in a platform like Langfuse or Laminar. Would this path work?

I think that could work - I'll give it a go. My goal is to monitor and prove effectiveness of steering/not using steering/changes to steering, skill use and MCP/tool use.

Let me know how it goes, I know other people have asked about audibility and it something I have wanted to dig in and study a bit more.