An autonomous RLM agent on rlm-harness that reads AI coding-agent session transcripts and their memory store, proposing a judgement-only distillation plan — what to prune, cross-reference, or promote into durable memory or a reusable Skill. - View it on GitHub