Phoenix Observability

This brings Phoenix's open-source observability platform into your Claude workflow for debugging and monitoring LLM applications. You get detailed tracing when something goes wrong, systematic evaluation runs across datasets, and real-time production monitoring without sending your data to a third party vendor. The skill has passed security audits from Gen Agent Trust Hub, Socket, and Snyk, which matters when you're instrumenting production systems. It's originally from the ai-research-skills collection and has solid traction with over 300 installs. Best for teams that want proper observability infrastructure but need to keep everything self-hosted, or anyone tired of flying blind when their prompts misbehave in production.