save-learnings

What problem does it solve?

Save learnings at the end of conversations across multiple persistence layers: the project memory, the mandatory wiki, the team repository, and the journal. It also auto-creates new skills, agent children files, and rules when repeating patterns are detected, removing the need for confirmation and enabling autonomous improvements.

Core Features & Use Cases

• End-of-session capture: manual invocation or marker-driven extraction to summarize and categorize learnings.
• Auto-create artifacts: new skills, rules, and children when applicable; updates agent memory and wiki accordingly.
• Wiki-first truth: mandatory wiki updates ensure current knowledge across the project, with optional doc-sync and journal logging.
• Team repo integration: changes are prepared, pushed, and summarized, enabling cross-team knowledge sharing.
• Marker-driven scope: parses <!-- learning --> sections to focus processing and reduce overhead.

Quick Start

Run /save-learnings at the end of a session or with --from-markers to automatically collect learnings into memory, wiki, and the team repo.