Prepare Agentness Arena v2 static Space

llm: final-review polish — openai max_tokens parity, errBody rename, rule-sync note, honest key placeholder, README inline list; rebuild standalone

llm(spectate): bind ENGINE explicitly, guard null mock move, add mock stage status lines (review follow-up)

llm(spectate): surface agentness verdict in spectate status line (app.js draws report on canvas + clobbers #hint each frame)

llm: browser spectate mode — LLM plays the human path (memPredict/humanMove); mock/anthropic/ollama

llm(player): guard NaN seeds + flag-as-value, trim .env keys, report nullAgentness count (review follow-up)

llm: headless batch CLI over runCellAsync (mock e2e pinned to runCell; JSONL out; .env keys)

llm(providers): provider-attributed errors on malformed/empty responses + baseUrl normalize (review follow-up)

llm: providers.js — anthropic/openai/ollama chat via bare fetch (UMD, injectable fetch)

docs(observe): note frozen-rival limitation in renderBundleText mirrors engine replay reconstruction (review follow-up)

llm: observe.js — replay/bundle rendering, induce/live/predict prompts, RULE parsing

llm(observe): clarify legend that grid digits are capped at 9; exact token values in tokens: line (review follow-up)

llm: observe.js — coordinate-labelled board text + MOVE parsing (UMD, shared node/browser)