Tinyworld Auto Batching
Developmenttinyworld-auto-batchingTinyworld Auto Batching is an Agent Skill for development published in the jasonkneen/tiny-world-builder repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Use when changing Tiny World Builder Auto palette behavior, model inference, suggestions, or provider/model/API-key logic.
QUICK START
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/jasonkneen/tiny-world-builder/blob/HEAD/.codex/skills/tinyworld-auto-batching/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/tinyworld-auto-batching, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Tiny World Auto Batching
Auto should not call the model on every placement.
Current intended model:
- Capture the current home board as sparse 8x8 JSON using
snapshotCells(). - Ask the model for a ranked batch of candidate tile actions matching
AUTO_SUGGESTIONS_SCHEMA. - Spend
autoSuggestionQueuelocally for multiple Auto placements. - Refresh after
AUTO_REFRESH_EVERYAuto placements or when the queue runs out. - Auto still places on the clicked cell. The model suggests tile/action definitions, not coordinates.
Implementation guardrails:
- Keep provider/model/key state shared with the Generate modal via
AI_LS. - Use
normalizeAutoActionbefore applying model output. - Use
adaptAutoSuggestionToCell(action, x, z)so same-kind clicks increase local intensity. - If no API key exists, open the Generate modal with the Auto key message.
- Do not add a backend or new dependency.
Validation:
- Stub
generateAutoSuggestionsin browser eval and verify two Auto suggestions only call the stub once. - Verify no network call is needed while spending cached suggestions.