learnself

What problem does it solve?

Helps agents stop wasting idle time by learning from past sessions and improves cross-agent continuity by handing unfinished work to the right next agent.

Core Features & Use Cases

• History-based self-learning: When an agent is idle, it scans prior transcripts, extracts repeating patterns and lessons, and saves learnings into memory for future performance.
• Automatic cross-agent handoff: Packages context (session, task, files, progress) and transfers it to another agent, tracking pending/accepted/completed/failed states.
• Session end work clearing: Consolidates unfinished tasks, prioritizes them, creates handoffs for what must continue, records status, and confirms to the human that the session is properly wrapped up.
• Multi-Agent Gang integration: Supports per-gang memory separation and optional sharing of learnings across gangs for better team-level improvement.

Quick Start

Ask an agent to run /learnself history to review what it has learned from recent past sessions.