| # Thousand-Token Terrarium — Codex Documentation Pack | |
| This archive contains a Codex-ready documentation set for the simplified top-down grid version of Thousand-Token Terrarium. | |
| ## Files | |
| - `AGENTS.md` — project rules and instructions for coding agents. | |
| - `docs/ARCHITECTURE.md` — product architecture and simulation design. | |
| - `docs/MEMORY_AND_EVENTS.md` — event log, memory, valence, rolling memory, beliefs, social memory. | |
| - `docs/LLM_PROMPTS_AND_SCHEMAS.md` — OpenRouter/Qwen policy, object compiler prompt, slow cognition prompt, validation rules. | |
| - `docs/RL_AND_FAST_POLICY.md` — heuristic/learned fast mind design, directive resolver, user-in-training guidance, rewards/eval. | |
| - `docs/EVAL_AND_TESTING.md` — frontend, backend, model, map, memory, and interaction testing expectations. | |
| - `docs/STAGES.md` — 5-stage development plan. | |
| - `docs/CODEX_USAGE_ORDER.md` — recommended order for using the docs. | |
| ## Intended use | |
| Copy these files into the root of the repo, then ask Codex to read `AGENTS.md` and begin Stage 1 from `docs/STAGES.md`. | |