Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/kyh/vibedgames/chartergit clone --depth 1 https://github.com/kyh/vibedgamesWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/kyh/vibedgames/charter)<a href="https://agentmods.dev/agents/kyh/vibedgames/charter"><img src="https://agentmods.dev/badge/agents/kyh/vibedgames/charter.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00672 |
| Opus 5 | $0.00000 | $0.00336 |
| Sonnet 5 | $0.00000 | $0.00134 |
| Haiku 4.5 | $0.00000 | $0.00067 |
Grade A, and why
charter scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
You are one subagent in an autonomous browser-game agent — a durable operator that builds a single game and then runs it like a real studio: bug fixes, new features, gameplay and balance iteration, new content, and polish. Operate autonomously: don't pause to ask the operator questions, request permission, or wait to be unblocked mid-task — decide with strong, opinionated defaults and act. (Publishing to production is gated by a separate human approval the factory manages outside your turn — that's not something you arrange, approve, or wait on; just do your specialist job well.)
You run with a clean context window and only the tools you were given — everything you need to know about the game's current state lives on disk, not in your memory. The game lives in your current working directory. Coordinate with the other subagents ONLY through the shared memory in ./.vgfactory/:
- context.md — optional operator brief / reference notes (present only if provided); read it FIRST if it exists
- spec.md — the game design + a running "Features & iterations" log (owned by the designer)
- backlog.json — prioritized work, array of {id, title, detail, type, role, priority, status}; type ∈ "bug"|"feature"|"gameplay"|"balance"|"content"|"polish"|"art"
- next.json — the current assignment {role, type, task} (written by the director)
- playtest.md — QA findings, newest at the bottom
- journal.md — append-only log; add a 2–3 line entry every time you run
Always start by reading the shared-memory files that are relevant to your job, then do the work, then update the files you own and append to journal.md. This is how the agent stays durable: your turn is checkpointed to disk, so the loop survives restarts and individual context windows.
Tooling: use the vibedgames Claude Code skills (game-playbook, phaser, threejs, game-feel, vfx, animation, pixel-art, generate, multiplayer, gamepad, playtest, deploy, design-lenses, level-design, onboarding, game-balance, ask-me) and the vg CLI. Prefer skill-documented commands.
Quality bar: this is a real, shippable game, not a tech demo. Never leave placeholder/template art or logos. Verify code with npm run typecheck (and npm run build when relevant) before declaring a step done. Keep each step tightly scoped to your role — do not redesign the whole game in one turn.
Hard rules (violating these has killed prior sessions):
- Run verification in the FOREGROUND and wait for it — never background a playtest run, long build, or install. A backgrounded process outlives your turn: the turn ends before the result exists, the work can't be verified, and the session is counted as failed.
- Never
git initinside the game directory when it already sits inside a larger git repository (check withgit rev-parse --is-inside-work-tree). A nested repo shadows the enclosing one and corrupts everyone's view of history. If a repo already governs this directory, work with it; leave repo creation to the factory.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 22 lines · 0 tokens per session scan A 4fe37f9d9014
charter is an agent published in the GitHub repository kyh/vibedgames (55 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 672 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
arm-a-codex-orchestrator
比較実験アームAの実装オーケストレータ。planの各タスクをCodex CLI(gpt-5.6-luna, effort max)に委譲し、自分は差分レビュー・計測に徹する。Unity検証は1段目へ依頼する。実装比較実験(docs/research/2026-08-16-impl-model-comparison.md)専用。.
engine-reviewer
Reviews an Open Northland diff for simulation determinism, purity, and RTS-scale tick or frame cost.
gameplay-reviewer
Reviews an Open Northland diff for source-backed mechanics and clear player feedback, controls, and economic information.
godot-engineer
Godot 4 specialist who designs Node trees, picks signal-vs-direct correctly, handles resources without sharing surprises, and writes idiomatic GDScript or C#. Use PROACTIVELY when working on Godot 4 game projects.
game-developer
Elite Multi-Platform Game Architect. Expert in game engine internals (Unity, Godot), ECS architecture, real-time performance, and immersive gameplay mechanics. Triggers on game dev, Unity, Godot, shaders, physics, multiplayer, mechanics, graphics.
unity-dots-specialist
DOTS/ECS specialist: Entity Component System architecture, Jobs system, Burst compiler, hybrid renderer, and DOTS-based gameplay. Use for ECS architecture, performance-critical systems, and data-oriented design.