jennydove/self-improving-claude

A framework for a self-improving Claude Code setup: it’s designed for a single product manager/operator who uses Claude daily and wants the setup to audit itself, fix drift, and accumulate knowledge across sessions. Core idea: two loops — a fast session loop (run every session via /done) that logs work, updates memory, detects new reusable skills, and flags skill improvements; and a slower quality loop (weekly/bi‑weekly audits) that verifies files against written standards. Key features: - Orchestrated sub‑skills (/done calls /ai-log, /update-memory, /check-for-skills, /update-skills, push config) - Audit commands (/audit-skills, /audit-claude-md) and fixes (/fix-skills, /fix-claude-md, /update-config) - Memory system (memory/MEMORY.md index + per-topic memory/*.md files) - Growth tooling to extract reusable skills - Staged automation playbook: manual -> scheduled audits -> full automation when trusted Scope: personal setup; future intent to support team setups.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.