Memory-centric self-improving harness for AI agents. A bidirectional growth loop — the cycle changes the agent and the human. ADRs, JSON schemas, and a dependency-free Python reference.
specifications adr self-improvement harness ai-agents claude agent-framework ai-agent llm prompt-injection anthropic llm-agent agentic-ai agent-memory claude-code memory-architecture self-improving-agent agent-rules security-by-absence agent-knowledge-cycle
-
Updated
Apr 10, 2026