What problem does it solve? When AI skill guidance is wrong, incomplete, or stale during Kotlin Multiplatform development, that knowledge is usually lost. This Skill captures each finding as a structured lesson file so corrections and better patterns flow back into the skills collection instead of being forgotten. ## Core Features & Use Cases - Structured Lesson Capture: Writes one markdown file per finding to docs/lessons/ with frontmatter fields for skill, date, severity, and type (correction, gap, better-pattern, deprecation, confirmation). - Deterministic Script: The create_lesson.py script generates date-stamped, slug-named files, never overwrites existing lessons, and inserts TODO placeholders for omitted fields. - Proactive Issue Drafting: For high-severity corrections, it drafts a GitHub issue for the upstream skills repo in the same turn, requiring explicit user confirmation before submission. - Use Case: After fixing a bug where a SharedFlow with replay=1 re-triggered navigation effects, run the script to record the correction with evidence so the kmp-mvi skill can be amended later by the harvester. ## Quick Start Ask the agent to write a lesson capturing what went wrong with a skill's guidance, including the correct pattern and evidence, and it will create the lesson file in docs/lessons/.
