session-intelligence-harvester

Harvest learnings automatically captures corrections, pattern drift, and fixes made during a session and routes them to your Reusable Intelligence Infrastructure components—CLAUDE.md, Constitution, and Agent files—so improvements persist across future sessions and mistakes don't recur.

Session harvesting identifies and encodes corrections made, format drift fixed, new patterns that emerged, and explicitly requested learnings, then classifies each insight and applies precise edits across multiple RII components to embed the learning permanently.

You can initiate harvesting manually by requesting it, or the Skill triggers automatically when corrections are detected, patterns emerge, or format issues are resolved during a session, then commits changes across your RII components.

Harvesting analyzes session corrections and patterns, determines their type and scope, then automatically routes insights to the appropriate RII components—whether that's documentation, rules, or agent behavior—ensuring learnings surface at the correct decision point.

After harvesting completes, changes are committed across multiple files in your RII, making learnings immediately available for future sessions so similar issues are prevented and patterns are recognized and applied consistently.