Open-source game operations kit

Keep shipped games alive.

MOStudio Kit gives Codex and Hermes an evidence-first operating layer for fragile, live, revenue-bearing game projects — from Unity clients and C++ servers to Lua contracts, player saves, release gates, and liveops incidents.

Evidence over confidence · BLOCKED is never PASS

studio session / route.log
you $client can't enter the game offline
reading project context + risk boundary
routeunity-client-offline-debugging
moderead-only diagnosis first
scopebootstrap / fallback / scene startup
Verified: artifact attached, exit 0
!BLOCKED: editor unavailable — not faked
$evidence > confidence
47canonical skills
22canonical agents
7installable packs
276/276routing evaluation
≠ PASSBLOCKED contract
Why this exists

Production changes the risk.

Once players and revenue depend on the project, a change can cross fragile clients, services, generated assets, databases, and player data. The kit treats that system as something to operate—not a blank canvas.

  • Protect player data and generated outputs.
  • Trace failures across the client, server, and pipeline.
  • Keep unavailable proof explicitly BLOCKED.
A clean greenfield game project contrasted with the complexity of operating a shipped live game.
The catalog

One operating model.
Seven studio lanes.

The router starts with the smallest matching workflow. Packs give teams a clean way to install only the surface they need.

01 / coreengine-agnostic
CORE

Studio Core

Intake, safety, evidence, collaboration, design, and governance for every project.

15 workflows
02 / clientunity
UNITY

Unity

Offline debugging, UI rendering, localization, asset integrity, batch builds, and scaffolding.

6 workflows
03 / backendmmorpg
NET

C++ / Lua MMORPG

Services, MySQL, protocol contracts, crash triage, authority review, and save migrations.

9 workflows
04 / operationsliveops
OPS

Production / LiveOps

Playtests, performance, economy, balance, release, stores, incidents, and telemetry.

8 workflows
05 / planningdelivery
PLAN

Production Management

Production planning, dependency risk, QA strategy, compatibility, and capacity verification.

5 workflows
06 / contentcreative
ART

Content Production

Level, narrative, art, animation, and audio production review workflows.

5 workflows
07 / productexperiments
DATA

Product Analytics

Product experimentation and governed live content rollout workflows.

2 workflows
Browse the workflows

Ask for the outcome.
Find the right route.

No slash commands or skill names required. Search by the failure, system, or gate you need to handle.

47 skills
No workflow matches that search. Try a system name such as Unity, Lua, MySQL, or release.
The trust contract

Trust is a data model, not a tone of voice.

Every meaningful claim carries a label. The kit makes it structurally difficult for an agent to turn missing evidence into a confident success report.

Verified

An observed command, test, build, primary source, or artifact supports the claim.

Snapshot

True for a named commit, version, configuration, or environment snapshot.

Unverified

A hypothesis or forecast that still needs evidence before it becomes a result.

BLOCKED

A runner, dependency, permission, live project, or required proof is unavailable.

CORE RULE · BLOCKED is never converted into PASS.
The studio lifecycle

Ask once.
Move through the right gates.

The catalog is not a checklist of 47 skills. Start with a desired outcome; the router selects a lane, then safety and evidence gates control how far the agent can go.

Lifecycle phases and outputs
  1. Request
    01Describe outcome
    natural language
  2. Intake
    02Scope + risk
    task packet
  3. Route
    03Diagnose / design
    smallest matching skill
  4. Change
    04Mutate with gates
    diff + restore path
  5. Verify
    05Run real checks
    command + exit + artifact
  6. Handoff
    06Report honestly
    verdict or BLOCKED
Read-onlyInspect and diagnose automatically.
Low riskVisible diff plus verification.
Medium riskExact scope, backup, restore path, reviewer.
High riskHuman approval and dry-run evidence.

Four common entry routes

You describe the job. The kit composes only the skills needed for that path.

Debug a failure

Intake → Debug → Verify

“Client can't enter offline mode. Fix it.”

studio-project-intakeevidence-first-debuggingsafe-project-mutationbuild-and-runtime-verification
Design a feature

Idea → Spec → Work packets

“Turn this mechanic into implementation work.”

game-feature-brainstorminggame-feature-to-specfeature-to-work-packetsreview-swarm
Adopt a project

Intake → Scaffold → Handoff

“Set this existing game repo up for agent work.”

studio-project-intakestudio-project-scaffoldbuild-and-runtime-verificationstudio-handoff
Ship or respond

Preflight → Gate → Observe

“Is this RC safe to ship?” / “Players found a dupe.”

Releaserelease-candidate-preflightstore-submission-checklist
Incidentliveops-incident-responsestudio-handoff

Human checkpoints remain in the loop. The router can choose a workflow automatically, but high-risk database, service-control, credential, destructive, and publish actions cannot cross the approval gate on their own.

Evidence from a skill-routed run on a live Unity MMORPG, showing verified fixes alongside blockers that remained blocked.
Field evidence

The contract survived a real project.

Two skill-routed runs shipped three fixes through their owning pipelines, preserved unrelated work, and kept two verification gaps honestly BLOCKED.

Verified · 32/32 config IDs · 34/34 combined sanitizer tests · 150 px content inside a 160 px UI bound · 0 foreign locks touched
Read the full case study →
Start with one command

Install once.
Ask for the outcome.

Choose your primary runtime, install MOStudio Kit from the stable GameStudio-CodexKIT repository, and describe the studio problem directly.

Hermesnpx skills add hoatv2211/GameStudio-CodexKIT -a hermes-agent -g -y
Codexcodex plugin marketplace add hoatv2211/GameStudio-CodexKIT
AskAudit this game project and route the work to the right studio skills.