Does this replace manual documentation like DISCOVERIES.md?

No, it is designed to complement manual logs. While DISCOVERIES.md is for major human-readable milestones, Session Learning provides a structured, agent-readable format for quick, reusable insights.

How does the skill decide what information to save?

It analyzes session transcripts for significant events such as solved error patterns, architectural decisions, and workflow improvements, typically focusing on issues that took more than 10 minutes to resolve.

Can I manually edit or search the stored learnings?

Yes, you can use the `/amplihack:learnings` command to show, search, or manually add learnings, or edit the YAML files directly in your editor.

Will searching through past learnings slow down my session start?

No, it uses a lightweight keyword-matching algorithm and a fail-safe design that ensures the session continues normally even if the injection process encounters an error.

Where are the learned insights stored?

Insights are stored locally in your project directory within `.claude/data/learnings/` in categorized YAML files, keeping your data under your control.

Session Learning

Name: Session Learning
Author: rysweet

byrysweet

•

生産性とワークフロー

Automates cross-session knowledge retention by extracting insights from coding transcripts and injecting relevant past solutions into new tasks.

概要

Session Learning is a specialized capability for the Claude Code environment that transforms transient session data into persistent, actionable knowledge. By analyzing session transcripts at termination, the skill extracts error patterns, workflow improvements, and architectural decisions into structured YAML files. When a new session begins, it uses a lightweight keyword-matching algorithm to surface and inject relevant historical context, ensuring the AI avoids repeating past mistakes and leverages previously discovered shortcuts without the need for manual documentation overhead.

主な機能

Automated insight extraction from session transcripts at session stop
Interactive command interface (/amplihack:learnings) for manual management and search
16 GitHub stars
Keyword-based relevance matching for automatic context injection at session start
Fail-safe design ensuring that learning processes never block or interrupt the primary coding workflow
Structured YAML storage across five specialized categories: errors, workflows, tools, architecture, and debugging

ユースケース

Recalling specific fixes for complex, recurring issues like circular dependencies or environment-specific bugs
Onboarding to a new project by reviewing a history of architectural trade-offs and design decisions
Scaling developer productivity by automatically documenting and surfacing non-obvious tool behaviors and shortcuts

概要

主な機能

Automated insight extraction from session transcripts at session stop
Interactive command interface (/amplihack:learnings) for manual management and search
16 GitHub stars
Keyword-based relevance matching for automatic context injection at session start
Fail-safe design ensuring that learning processes never block or interrupt the primary coding workflow
Structured YAML storage across five specialized categories: errors, workflows, tools, architecture, and debugging

ユースケース

Recalling specific fixes for complex, recurring issues like circular dependencies or environment-specific bugs
Onboarding to a new project by reviewing a history of architectural trade-offs and design decisions
Scaling developer productivity by automatically documenting and surfacing non-obvious tool behaviors and shortcuts