microagent: align runtime SYSTEM_PROMPT with training data (multi-line bash blocks) — removes scaffold drift between SFT and inference

microagent: tighten for TB2 eval (turns 35->20, parse_fails 4->3, history 4->5, tokens 1024->1536, drop enable_thinking kwarg, expand long-cmd list)

Add INSTANCE_RUNBOOK.md for Claude-on-instance briefing

Add live progress callback (ETA, tok/s, GPU mem, loss EMA, cost)

Upload data/pipeline_v2_log.txt with huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload README.md with huggingface_hub

initial commit