self-improving

Self-learning system for Claude Code that captures corrections and updates CLAUDE.md automatically

Intelligent command history tracking with automatic failure detection. Tracks all bash commands Claude runs (successful and failed) with semantic directory aliasing and command mapping.

Self-improving learning system that detects friction signals (corrections, repetitions, tool failures), extracts improvement candidates, and proposes rule/skill updates with explicit approval workflow.

Detects correction or long-term note-taking scenarios, runs lightweight scholar preflight on fresh tasks, keeps the agent in an explicit closed loop until the user confirms completion, escalates into Ascended Mode after repeated failed corrections or manual trigger, archives detailed `mistake` / `note` entries into project/global stores, and refreshes cache-like distilled memory.

Continuous learning hooks for gladiator MCP (observe patterns, reflect on them)

Session feedback analysis - capture skill bugs, enhancements, and positive patterns as GitHub issues