What file format does this skill require?

It is specifically designed to parse the .jsonl files automatically generated by Claude in the .claude-trace directory.

Does this skill help reduce AI costs?

Yes, by identifying high token consumption patterns and inefficient tool calls, you can optimize your workflows to be significantly more cost-effective.

Can I compare multiple sessions at once?

Yes, the 'compare' action allows you to aggregate metrics across several trace files to identify performance trends and usage anomalies over time.

Is this useful for debugging tool failures?

Absolutely. It categorizes errors like rate limits, timeouts, and execution failures, providing recommendations on how to resolve recurring technical bottlenecks.

How does this differ from the /transcripts command?

While /transcripts shows the conversation content, this skill analyzes the API-level JSONL data for technical metrics like latency, token usage, and specific tool execution errors.

Claude Trace Analytics

Name: Claude Trace Analytics
Author: rysweet

byrysweet

•

分析と監視

Analyzes Claude Code trace files to provide actionable insights into session health, token consumption, and tool performance.

Claude Trace Analytics is a specialized diagnostic skill designed to decode and interpret the underlying JSONL trace data generated during Claude Code sessions. Unlike standard transcripts that focus on conversation history, this skill drills into API-level metrics, allowing developers to diagnose latency issues, identify recurring error patterns, and optimize token efficiency. By providing deep visibility into tool execution frequencies and model interactions, it helps users refine their agentic workflows, troubleshoot failures, and manage operational costs effectively.

主な機能

01Automated error pattern detection and categorization

0217 GitHub stars

03Comprehensive token usage and efficiency reporting

04Tool usage optimization and bottleneck identification

05JSONL trace file parsing and automated health scoring

06Comparative performance analysis across multiple sessions

ユースケース

01Debugging slow or failing Claude Code sessions to find technical root causes

02Auditing token expenditure and optimizing prompt efficiency to reduce costs

03Analyzing tool execution patterns to identify parallelization opportunities

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rysweet/amplihack session-replay

For use in Claude.ai and ChatGPT