MOStudio Kit — AI Agent Skills for Unity and MMORPG Operations
Evidence-first workflows for operating live games with Codex and Hermes Agent.
MOStudio Kit is a catalog of AI agent skills for Unity clients, C++ and Lua MMORPG servers, game databases, content pipelines, releases, and live operations. It helps an agent diagnose failures, plan bounded changes, protect player data, and prove what was actually verified. It is built for maintaining shipped games—not generating a new game from scratch.
MOStudio Kit is distributed from the GameStudio-CodexKIT repository. The
public product name is MOStudio Kit; repository URLs and the stable plugin ID
game-studio-codex-kit keep the technical project name.
Why MOStudio Kit
- Built for live game operations. Workflows cover fragile clients, authoritative servers, generated assets, player data, builds, incidents, and releases.
- Evidence instead of confidence. Results distinguish
Verified,Snapshot,Unverified, andBLOCKED. An unavailable runner never becomes a fake PASS. - Safe changes by default. Read-only diagnosis comes first. Higher-risk work requires visible scope, review, backup or restore evidence, and human approval where appropriate.
- One canonical catalog. Codex and Hermes Agent use the same workflows, evidence contracts, and risk boundaries.
Visual overview
These seven panels summarize the product before the detailed guides: what the kit operates, how requests are routed, and why evidence and mutation gates matter in a live game project.
What you can do
Describe the outcome in normal language. The router selects the smallest matching workflow; you do not need to memorize skill names.
| Request | Typical workflow |
|---|---|
| “Audit this Unity/MMORPG project and identify the right workflow.” | Project intake and workspace routing |
| “The Unity client cannot enter offline mode.” | Offline bootstrap debugging |
| “The C++ game server crashed; here is the stack trace.” | Build-bound crash triage |
| “Check whether this Lua client action is validated server-side.” | Protocol and server-authority review |
| “Plan this MySQL schema change without risking player data.” | Dry-run migration safety |
| “Prepare a release go/no-go decision with rollback evidence.” | Release candidate preflight |
The catalog also covers Unity UI and asset integrity, localization authority, performance budgets, telemetry contracts, game economy review, playtests, LiveOps incidents, store submission, production planning, and governed multi-agent work.
Install
Choose one primary runtime. A repository clone is not required for normal use.
Codex App and CLI
codex plugin marketplace add hoatv2211/GameStudio-CodexKIT
codex
Open /plugins, select the MOStudio Kit marketplace, install the plugin,
and start a new task or CLI session.
Hermes Agent
npx skills add hoatv2211/GameStudio-CodexKIT -a hermes-agent -g -y
Start a new Hermes Agent session after installation. See the English installation and project activation guide for upgrades, removal, project-local installation, and governed scaffolding.
Quick start
Give the agent a goal, readable and writable scope, protected paths, and the evidence you expect:
Repository: D:/Games/MyMMO
Goal: determine why the Unity client cannot enter offline gameplay.
Readable scope: Assets/, ProjectSettings/, Logs/.
Writable scope: none until the root cause is verified.
Do not touch: .env, databases, generated localization, production services.
Return: selected workflow, ranked evidence, commands, exit codes, artifacts,
limitations, and explicit Verified/Snapshot/Unverified/BLOCKED results.
For more examples, use the Vietnamese skill and agent guide or the complete English skill and agent guide.
Evidence, not confidence
The current catalog contains 49 canonical skills, 24 canonical agent roles, seven installable packs, and 306 deterministic eval cases for Tier-A routing. Local deterministic tests, model-runner evidence, real-project dogfood, and lifecycle maturity are deliberately reported as separate things.
Some workflows remain experimental, and missing runtime, project, approval,
or session-history evidence remains BLOCKED. Read the
current adoption evidence for observed results and known
limitations; do not infer release readiness from badges alone.
Documentation
| Guide | Purpose |
|---|---|
| Hướng dẫn Skill và Agent bằng tiếng Việt | Cách chọn và sử dụng toàn bộ skill/agent với ví dụ thực tế |
| English Skill and Agent User Guide | Wiki-ready user guide, Golden Paths, prompts, agents, installation, and end-to-end examples |
| Skill Catalog | All skills, routing triggers, risk levels, and output artifacts |
| Architecture | Source authority, routing, distribution, evidence, and safety model |
| Project Adoption Workflow | Report-only intake, role-first routing, and governed project scaffolding |
| Adoption Evidence | Verified results, snapshots, limitations, and lifecycle blockers |
| Skill Authoring Guide | Rules for contributors changing canonical workflows |
The GitHub Pages landing page provides the visual overview. Detailed adapter, evaluation, mutation, archive, and maintenance procedures live in the linked documentation rather than this README.
Project status
The catalog is beta based on maintainer-confirmed studio adoption. Individual skills may be beta or experimental, and each task must still earn its own evidence. Codex App/CLI and Hermes Agent are the primary distribution targets.
Repository maintenance uses Python 3.11 and the deterministic gates documented in Adoption Evidence. Issues and proposals can be opened in GitHub Issues.
License
MIT © MAD Studio. Contributions to skills, agents, documentation, or evaluation fixtures should follow the Skill Authoring Guide and repository Operating Contract.
No comments yet
Be the first to share your take.