common-session-retrospective

To auto-improve agent skills from session corrections, run a retrospective process that identifies correction signals, classifies root causes, and determines targeted updates to existing skills or workflows. The system extracts correction signals to reveal skill gaps, then syncs changes across all agent skill directories and refreshes the skills index automatically.

When analyzing conversation corrections for skill gaps, root causes are classified into five categories: Skill Missing, Incomplete, Example Contradicts Rule, Workflow Gap, and Trigger Miss. This classification determines targeted updates to existing skills, references, or new workflows to close the detected gaps.

You need to run a retrospective to detect skill gaps after any session involving user corrections, rework, or retrospective requests. It detects signals in sessions where outputs were corrected or requests for improvement were made, ensuring continuous skill library enhancement.

Yes, you can update multiple agent skill directories and regenerate the skills index automatically. The process applies proposed changes across all agent skill directories and regenerates the skills index, ensuring consistent updates to references and new workflows following root cause analysis.

The best way to extract correction signals from automation sessions is using a retrospective analysis that identifies rework or improvement requests. It classifies the root cause of each correction and proposes targeted updates to existing skills, references, or new workflows to systematically close the revealed gaps.