Generates evidence-based performance reports on AI agents and skills by analyzing Claude Code session transcripts.
Field Report is a sophisticated diagnostic tool designed for developers to audit and improve the performance of their Claude Code plugins, skills, and agents. By analyzing conversation logs and tool traces, it produces structured, narrative evaluations that highlight task completion rates, instruction adherence, and tool efficiency. It prioritizes data privacy through automatic sanitization while providing actionable recommendations to help maintainers refine AI behavior based on real-world evidence from specific session IDs or natural language history.
主要功能
01Automated session discovery via natural language or explicit session IDs
02Privacy-safe data sanitization to redact sensitive tokens and PII
033 GitHub stars
04Structured Markdown report generation with actionable developer recommendations
05Evaluation of conversation efficiency and error recovery latency
06Evidence-based analysis of instruction adherence and tool usage patterns
使用场景
01Optimizing tool usage sequences to improve session speed and reliability
02Auditing instruction adherence for custom AI agents before deployment
03Debugging why a specific skill or tool failed during a complex workflow
