lesson-extractionlisted

# Lesson Extraction ## Overview Every completed task contains lessons that can prevent future mistakes or accelerate future work. Extraction captures these lessons in a structured format that makes them searchable, scorable, and expirable. The goal is not to document everything. It is to capture lessons that are specific enough to be actionable, general enough to be reusable, and evidence-based enough to be trustworthy. ## The Iron Law ``` ALL 4 QUALITY FILTERS MUST PASS — NO PARTIAL CREDIT ``` The learning playbook is only valuable if entries are actionable and generalizable. Without quality filters, it accumulates vague observations ("tests are important"), project-specific trivia, and gut feelings that dilute the signal. Each filter catches a different failure mode — specificity prevents vagueness, generalizability prevents overfitting, actionability prevents philosophy, and evidence prevents speculation. A candidate that fails ANY filter (specific, generalizable, actionable, evidence-based) is discarded. Better to extract 2 strong lessons than 10 weak ones. ## The Process ### Step 1: Identify Candidates Review the completed work and look for: - **Mistakes made**: What went wrong and why - **Non-obvious solutions**: Fixes that required investigation to find - **Patterns discovered**: Conventions or approaches that worked well - **Time sinks**: Steps that took longer than expected and why - **Corrections received**: Feedback from the user that changed approach #