Do I need a Claude API key to use this skill?

The 'Analyze' mode requires the Claude CLI to be configured, but the 'Stats Only' mode works locally on your session logs without additional API calls.

Where does the tool find my Claude Code logs?

It automatically scans the default ~/.claude/projects/ directory where the Claude CLI stores session history.

What does a score of 100 represent?

A perfect score indicates a session with zero waste, where the AI used the minimum necessary tokens and steps to complete the user's request accurately.

How does Session Evaluator help reduce API costs?

It identifies redundant file reads and unnecessary context bloat, helping you refine your prompts to minimize token consumption.

Session Evaluator

Name: Session Evaluator
Author: xtone

byxtone

•

分析与监控

Analyzes and scores Claude Code session efficiency to reduce token waste and optimize AI development workflows.

Session Evaluator is a specialized web-based tool designed to audit Claude Code interactions by analyzing session logs (JSONL). It evaluates AI behavior across seven key dimensions—including token efficiency, task relevance, and error recovery—to provide a detailed performance score out of 100. By identifying redundant file reads, unnecessary context bloat, and suboptimal model selections, this skill helps developers refine their AI configurations and prompts to achieve leaner, more cost-effective, and faster development cycles.

主要功能

01Detailed token analysis including input volume and cache utilization tracking

02Web-based dashboard for browsing and selecting Claude Code projects and sessions

03Dual-mode analysis: Full AI evaluation or lightweight 'Stats Only' mode

047-point scoring system focused on efficiency and waste reduction

05Actionable improvement suggestions categorized by hooks, rules, and skills

063 GitHub stars

使用场景

01Auditing AI usage to minimize API costs by identifying token-heavy redundancies

02Optimizing complex development tasks to ensure sub-agents aren't duplicating work

03Refining custom instruction files like CLAUDE.md based on historical session performance

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xtone/ai_development_tools evaluate-session

For use in Claude.ai and ChatGPT

主要功能

01Detailed token analysis including input volume and cache utilization tracking

02Web-based dashboard for browsing and selecting Claude Code projects and sessions

03Dual-mode analysis: Full AI evaluation or lightweight 'Stats Only' mode

047-point scoring system focused on efficiency and waste reduction

05Actionable improvement suggestions categorized by hooks, rules, and skills

063 GitHub stars

使用场景

01Auditing AI usage to minimize API costs by identifying token-heavy redundancies

02Optimizing complex development tasks to ensure sub-agents aren't duplicating work

03Refining custom instruction files like CLAUDE.md based on historical session performance

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xtone/ai_development_tools evaluate-session

For use in Claude.ai and ChatGPT