What is the Langfuse Trace Analysis skill?

It is a specialized capability for Claude Code that helps developers debug LLM applications by analyzing Langfuse traces and connecting those traces to the underlying codebase for faster troubleshooting.

Can it help improve prompt quality?

Yes, by using the 'prompts' mode, you can inspect the specific prompt templates and variables used at each step of a trace to identify logic errors or quality issues.

Do I need other plugins for this to work?

This skill utilizes the data-retrieval scripts from the mberto-compound collection to fetch trace information before performing the analysis.

How does the report generator help?

It creates a standardized summary including the symptom, root cause, trace evidence, and a recommended fix, ensuring consistent documentation and faster resolution of bugs.

Can I analyze latency issues with this skill?

Yes, the skill includes a specific 'flow' mode designed to visualize and analyze execution timing, making it easy to spot latency bottlenecks in multi-step LLM chains.

Langfuse Trace Analysis

Name: Langfuse Trace Analysis
Author: mberto10

bymberto10

Analytics & Monitoring

Diagnoses and resolves LLM workflow issues by performing structured root-cause analysis on Langfuse traces.

About

Langfuse Trace Analysis is a specialized skill for Claude Code designed to streamline the debugging and optimization of LLM-powered applications. By integrating Langfuse observability data directly with codebase investigation, it allows developers to quickly classify symptoms like latency, data gaps, or quality issues and map them to specific code segments or prompts. The skill automates the retrieval of execution flows and I/O data, providing a structured framework to identify why a workflow failed and generating concise reports with actionable fixes to bridge the gap between production logs and source code.

Key Features

Multi-mode trace retrieval for I/O, execution flow, and prompt inspection
Automated report generation with evidence-based fix recommendations
Automated symptom classification for data gaps, latency, and quality issues
Pattern-based investigation heuristics for common LLM failure modes
0 GitHub stars
Structured root-cause analysis linking trace evidence to specific code files

Use Cases

Identifying latency bottlenecks in complex multi-step chains or RAG pipelines
Analyzing prompt quality and input/output mismatches at failing steps
Debugging why an LLM agent failed to call a specific tool or API

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mberto10/mberto-compound langfuse-trace-analysis

For use in Claude.ai and ChatGPT

Download Skill

GitHub