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 skills add portals-labs/portals-plugin-claude --skill portals-server-simgit clone --depth 1 https://github.com/portals-labs/portals-plugin-claudeWrote 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/skills/portals-labs/portals-plugin-claude/portals-server-sim)<a href="https://agentmods.dev/skills/portals-labs/portals-plugin-claude/portals-server-sim"><img src="https://agentmods.dev/badge/skills/portals-labs/portals-plugin-claude/portals-server-sim/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/portals-labs/portals-plugin-claude/portals-server-sim"><img src="https://agentmods.dev/badge/skills/portals-labs/portals-plugin-claude/portals-server-sim.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00101 | $0.01734 |
| Opus 5 | $0.00051 | $0.00867 |
| Sonnet 5 | $0.00020 | $0.00347 |
| Haiku 4.5 | $0.00010 | $0.00173 |
Grade A, and why
portals-server-sim 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 11d 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.
How it starts
The opening of the file, as written. The whole thing — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Server-simulated games
The portals-server-scripts skill covers the contract — a root server.js that Portals runs as an invisible authoritative participant. This is the next level: the server script simulates the game itself, stepping the same physics the clients ship and broadcasting the result, so nobody's tab owns the world.
Read references/server-sim.md for the full pattern with the worked FUTBOL example — protocol, tick, prediction, fallback ladder, measured budgets, and the virtual-clock test harness. It is the official documentation, copied verbatim.
The default for multiplayer games
If the game is multiplayer, build it server-simulated. Host-authoritative multiplayer (one tab simulates, the rest follow) is not the starting point — it is the fallback rung a server sim keeps underneath for the moments the server is away. The server sim buys fairness (games are continuous and contested more often than they first look — a shared ball, ramming, racing, even a grabbed pickup — and a host's zero self-latency is a real competitive advantage), witnessed outcomes rather than claims, and symmetry — the server sits at the relay, ~25 ms from everyone on a regional channel.
The exceptions are narrow: purely cosmetic shared spaces (a co-op drawing canvas, an emote wall) and slow turn-based games whose pace hides latency, with a plain server script as referee where anything needs deciding. When in doubt, take the server sim — retrofitting authority into a shipped host-authoritative game is much harder than building on it, while the host-authority fallback comes with the server sim for free.
The rules that make or break it
- One simulation, shared byte-identically. The client and server must run the same game rules from one module. A ball simulated two slightly different ways is a desync nobody can debug.
- The sim module stays plain. Pure data in, pure data out (
{ x, z }fields, no THREE.js, no DOM, no clock of its own). Side effects go through hooks so each authority decides what a goal or a sound means. - Integration must be step-rate independent. The server ticks at 20 Hz against the sandbox's 50 ms timer floor; clients render at 60–120 Hz. Solve movement exactly over
dt(vel = settle + (vel - settle) * Math.exp(-dt * k)) instead of accumulating per-frame adds, or the two ends settle at different speeds and prediction argues with its authority for as long as a key is held. - The sandbox takes one file — no
import, norequire— so share at build time. Bundle the referee entry plus the sim module intodist/server.jswith esbuild (format: "iife", unminified so it reads next to its logs); that bundle is what a push ships. - The server marks its snapshots. One field (
sv: 1) is the entire handshake: the first marked snapshot flips every tab, host included, into a follower; when the marked stream goes quiet the host resumes simulating. One protocol, two places it can run. - Inputs go up on change, not on a metronome — no faster than every 60 ms, and every 100 ms regardless so a held key stays alive. Carry a sequence number. On the server an input drives whatever seat the team sheet assigned that player, never whatever the packet asks for, and a stale input stops driving its actor after ~700 ms.
- The clock ticks on the server, never as a published deadline — a deadline is read against each player's own clock, and a wrong device clock would show a wrong match.
- Predict yourself, dead-reckon everyone else, replay contested moments from events. Ease your own actor onto the authority's version rather than snapping. Dead-reckon others on the sim's own drag constants — extrapolating at snapshot velocity in a straight line walks a released actor well past where the physics stopped it. Never predict a contested outcome; play it when
evsays so. - Subtract the ocean. A snapshot is already a round trip old. Echo the last acted-on sequence number in each actor's row; the gap between sending
qand reading it back — both ends off the guest's own clock — is the whole loop. Project the ball and opponents forward by one downlink, your own actor by the full trip. - The fallback ladder must stay walkable. A server sim that takes the game down with it has failed. Three rungs, each a working game: server simulates → server referees only while the host simulates → no server at all (peer-run, which is the ordinary state of every unpublished game). Silence is the signal: guests fall back after a beat and a half, the host starts simulating half a beat later, both measured from the same last-marked snapshot so guests are listening before the host talks.
- Boot resumes shared state. Shared state outlives the script. A hot-swapped server reads its own
server:*keys back, prunes them against the live roster, and carries on — being swapped mid-match must not hand anybody a fresh 0-0. - Coalesce state writes. Scoreboard once a second, directory-style state on a ~120 ms timer; the smooth clock rides the snapshots. Keep every value inside 8 KB: packed arrays over objects, two-decimal rounding, capped list sizes.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 11d ago First seen · 48 lines · 101 tokens per session scan A 44d4336890a4
portals-server-sim is a skill published in the GitHub repository portals-labs/portals-plugin-claude (4 stars, last pushed 2d ago), licensed ISC. It adds 101 tokens to every session and 1,734 once invoked, about $0.0005 per session on Opus 5. 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 2026-08-31.
Other skills, from other repositories
gameobject-component-destroy
Destroy one or more Components from a target GameObject. Missing (null) components are skipped — they cannot be destroyed. Use 'gameobject-find' and 'gameobject-component-get' to identify the components first.
unity-version-split
Split a C# file into Unity 6.5+ and pre-Unity 6.5 variants. Use when a file needs different implementations for different Unity versions due to API changes (e.g., EntityId vs int, GetEntityId vs GetInstanceID).
godot-signals-groups
Build event-driven, decoupled Godot 4.7 gameplay with signals and node groups: declare and emit custom signals, connect with Callables (incl. bind/one-shot), and broadcast to many nodes via groups and callgroup. Use when wiring node communication in a Godot project, replacing tight references with signals…
unity-addressables
Manage Addressables groups, entries, profiles and content builds (com.unity.addressables, reflection-based).
motion
How an agent turns a character mesh into a usable animated FBX — and how to judge whether the result is shippable.
threejs-exposure-color-grading
Build a measured exposure and grading path in Three.js. Use for a 64x36 encoded luminance meter, asynchronous readback, weighted log-average exposure, asymmetric adaptation, single tone-map ownership, and a generated 32-cube post-tone-map LUT.