How does it manage token consumption?

It uses a custom preprocessor to sample only relevant parts of logs, such as errors and session boundaries, significantly reducing the data sent to the LLM.

Can I filter logs by date?

Yes, the --since argument allows you to process logs from specific timeframes, such as '7d' for the last week or '24h' for the last day.

Does this skill protect my sensitive data?

Yes, the preprocessor applies redaction patterns to logs and truncates large fields before analysis. Raw logs are never committed to your repository.

What are the main outputs of this skill?

It generates two primary files: a human-readable lessons.md for manual review and a machine-readable lessons.jsonl for programmatic integration.

Why does the skill sometimes skip regeneration?

To prevent noise and file churn, the skill uses 'Run Scope' logic that skips updating the lessons files if fewer than five new sessions are detected.

Lessons Extractor

Name: Lessons Extractor
Author: BryceEWatson

byBryceEWatson

0•

学習とドキュメンテーション

Extracts and organizes reusable insights, tool-usage patterns, and debugging solutions from Claude Code session logs.

Lessons Extractor is a specialized utility for Claude Code that automatically synthesizes project experiences into a structured knowledge base. By analyzing local session logs, it identifies successful workflows, recurring tool failures, and platform-specific gotchas—such as Windows compatibility issues—outputting them into both human-readable Markdown and machine-readable JSONL. This allows developers and teams to build a persistent 'memory' of best practices, preventing the repetition of past mistakes and optimizing future AI interactions through data-driven insights.

主な機能

01Smart 'Run Scope' logic that prevents file churn by only updating during substantial activity.

02Extraction of tool failures as first-class data to improve future error recovery strategies.

03Automated log discovery with cross-platform support for local Claude Code projects.

04Automated categorization of lessons into workflow, debugging, and architecture patterns.

05Intelligent event sampling and truncation to minimize token usage and costs.

060 GitHub stars

ユースケース

01Creating an automated 'Lessons Learned' log for team retrospectives and onboarding.

02Building a project-specific knowledge base of coding standards and successful AI patterns.

03Debugging recurring environment issues by analyzing aggregated tool failure logs.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bryceewatson/bryce-labs-toolkit lessons-extractor

For use in Claude.ai and ChatGPT

Download Skill