Pushing general-purpose LLMs to their chess limits — no fine-tuning, no specialized models. Pure prompt engineering, RAG, chain-of-thought, and multi-agent orchestration. A scientific framework for measuring how far inference-time techniques can lift raw LLM chess performance.
21 files for Claude Code, Codex and OpenCode: architecture-boundary-review, audit-assistance-provenance, bootstrap-workspace, build-run-bundle and 17 more — 1,161 tokens loaded in every session.
AGENTS.md A 1,161 tok .agents/skills/architecture-boundary-review/SKILL.md A 33 tok .agents/skills/audit-assistance-provenance/SKILL.md A 40 tok .agents/skills/bootstrap-workspace/SKILL.md A 46 tok .agents/skills/build-run-bundle/SKILL.md B 30 tok .agents/skills/code-review/SKILL.md A 35 tok .agents/skills/database-migration/SKILL.md A 30 tok .agents/skills/design-research-experiment/SKILL.md A 40 tok .agents/skills/implement-chess-environment/SKILL.md A 38 tok .agents/skills/implement-decision-strategy/SKILL.md A 35 tok .agents/skills/implement-domain-contract/SKILL.md A 31 tok .agents/skills/implement-evaluator-metric/SKILL.md A 33 tok .agents/skills/implement-knowledge-packet/SKILL.md A 36 tok .agents/skills/implement-multimodal-observation/SKILL.md A 32 tok .agents/skills/implement-provider-adapter/SKILL.md A 34 tok .agents/skills/implement-runtime-state-machine/SKILL.md A 29 tok .agents/skills/prepare-human-decision-gate/SKILL.md A 33 tok .agents/skills/release-readiness/SKILL.md A 39 tok .agents/skills/test-fault-replay/SKILL.md A 35 tok .agents/skills/update-literature-map/SKILL.md A 35 tok .agents/skills/write-or-update-adr/SKILL.md A 41 tok These files are maelrx/Zugzwang's own configuration — they tell Claude Code, Codex and OpenCode how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.