update-agent-learnings

You can document session learnings by extracting actionable patterns and mistakes, then propagating them to a central learnings repository. This skill categorizes insights into Global, Claude Code Only, or Agent-Specific scopes to ensure uniform behavior improvements across all subagents.

The best way to synchronize agent learnings is using a phased workflow that updates a central file and per-agent files. This process analyzes the session, extracts learnings, updates the central repository, and synchronizes agent files with validation to ensure alignment.

Agent learning propagation works by applying extracted insights to defined scopes: Global, Claude Code Only, or Agent-Specific. Updates are pushed to the central repository and then synchronized to either all subagents or the targeted agent file accordingly.

Yes, you can capture conversation insights and apply them to Claude Code agent files. The skill updates the central learnings repository and synchronizes the relevant per-agent files, ensuring the Claude Code scope receives the extracted behavioral patterns.

You should run a session analysis to extract agent learnings immediately after a challenging session. This timing allows you to capture concrete patterns, mistakes, or insights while the context is fresh, propagating them to the central file and relevant agent files for validation.

Session learnings remain siloed across agents when there is no centralized capture mechanism to propagate updates. Without a standardized workflow to extract insights and synchronize per-agent files, behavioral improvements fail to align across the entire subagent network.