Orchestrator-first Codex skills for turning raw game ideas into validated gameplay packages, Chinese production-ready game specs, agent-executable execution plans, and dispatch-ready worker packets.
简体中文 · FAQ · Licensing · Commercial Offers · Release Notes
## What This Is This repository packages a game-design workflow for Codex around one hard rule: `gameplay-design-orchestrator` is the only normal upstream entry point. Instead of exposing ten unrelated prompts and hoping the user manually glues them together, this suite: - locks direction before system expansion - drives the gameplay stages in order - keeps scope discipline visible - compiles the result into a downstream-ready gameplay package - expands that package into a full Chinese game spec when needed - compiles that spec into an execution plan for agent-driven implementation - includes a first runner prototype for stepping through that execution plan - includes a dispatch layer that emits stable worker packets for downstream execution - includes a runner-managed `dispatch -> ack -> complete` lifecycle so packet export and worker ownership are separate steps - includes dispatch review-gate checks so completion evidence must match the active dispatch and acknowledged worker - includes a worker adapter so external workers can execute against a dispatch directory instead of raw runner internals The result is not just "better ideation." It is a more reliable bridge from vague concept to structured execution. ## Why This Repo Is Different Most public prompt packs for game design fail in one of three ways: | Typical Prompt Pack Failure | What This Suite Does Instead | | --- | --- | | Starts with theme, lore, or mechanics soup | Starts with direction lock, player promise, and constraints | | Treats every prompt as standalone | Treats downstream stages as a controlled chain under the orchestrator | | Produces pretty but non-executable prose | Produces package structure, registries, spec-ready outputs, and execution-ready task graphs | | Has no validation | Includes local validators and a real example | | Expands scope too early | Forces MVP keep/cut decisions before downstream handoff | ## Included Skills | Skill | Role | | --- | --- | | `gameplay-design-orchestrator` | Upstream controller for direction lock, ordered stage execution, and package assembly | | `gameplay-input-normalizer` | Converts raw input into stable constraints, audience pressure, platform limits, and emotional target | | `gameplay-fantasy-loop-designer` | Defines player promise, identity fantasy, core action, and loop structure | | `gameplay-pacing-and-structure` | Designs time-layer pacing, early highs, frustration points, and replay pull | | `gameplay-world-and-narrative-weaver` | Wraps gameplay in world rule, conflict, and act pressure without replacing the loop spine | | `gameplay-system-weaver-and-scope-cutter` | Maps systems, resources, MVP keep/cut, and production danger zones | | `gameplay-coding-handoff-compiler` | Compiles the locked package into scenes, UI, objects, states, variables, and prototype acceptance | | `game-design-spec` | Expands the mature gameplay package into a full Chinese game design spec and task pack | | `game-design-execution-compiler` | Compiles the mature spec into `execution-plan.json` and `execution-plan.md` for agent execution | | `game-design-execution-runner` | Drives one execution-plan task at a time, persists state, prepares bounded worker handoffs or dispatch packets, tracks `dispatch -> ack -> complete`, enforces dispatch-bound review gates, and exposes a worker adapter interface | For topology details, see [docs/skill-catalog.md](docs/skill-catalog.md). ## Pipeline ```text Raw game idea -> gameplay-design-orchestrator -> gameplay-input-normalizer -> gameplay-fantasy-loop-designer -> gameplay-pacing-and-structure -> gameplay-world-and-narrative-weaver -> gameplay-system-weaver-and-scope-cutter -> gameplay-coding-handoff-compiler -> Gameplay Design Package -> game-design-spec -> Full Chinese Game Design Spec -> game-design-execution-compiler -> execution-plan.json + execution-plan.md -> game-design-execution-runner -> execution-run-state.json + worker handoff / dispatch packet ``` Operational rules: - Do not treat the six downstream `gameplay-*` stage skills as normal free-entry ideation tools. - Once the formal chain starts, all six downstream stages are mandatory in order. - Only allow isolated stage revision when a package already exists and the user explicitly requests stage-level rework. ## Proof: Included Example This repository includes a fully validated gyro-battle web game example: - Index: [examples/gyro-battle/00-direct-output-index.md](examples/gyro-battle/00-direct-output-index.md) - Gameplay package: [examples/gyro-battle/final-package/](examples/gyro-battle/final-package/) - Full spec: [examples/gyro-battle/final-spec/](examples/gyro-battle/final-spec/) - Execution plan: [examples/gyro-battle/final-execution-plan/](examples/gyro-battle/final-execution-plan/) - Case study write-up: [docs/gyro-battle-case-study.md](docs/gyro-battle-case-study.md) The example package includes: - direction lock between two competing concepts - loop-back notes from pacing and scope pressure - a gameplay package with assumption list, system indexes, formulas, and prototype acceptance - a full spec with system rules, UI tasking, balance tasks, art/audio constraints, QA, and delivery mapping - an execution plan with small dependency-aware tasks tied back to source refs and canonical IDs - a runner-ready path that can initialize state, emit worker dispatch packets, acknowledge worker pickup, and step through tasks in dependency order This is a complete validation artifact, not just a visual showcase. ## Visual Preview
