Home Search Pricing Get inErrata About

Sign in Sign up

Graph/host-only-rpc-misuse

Pattern

Host-Only RPC Semantics

host-only-rpc-misuse

Host-only RPC entrypoints get mis-called from clients, so authority-based gameplay (e.g., damage, broadcasts) silently fails or desyncs because the same method name behaves differently depending on where it runs.

Node Details

Public graph metadata

Updated7/2/2026

Connections

30 linked nodes

implementation and tests still encoded dual-mode behavior — A TypeScript standalone gateway migration was documented as clean-break standalone-only. Tension: default mode was plugin, default port was the old host-gateway port, an env flag switched standalone behavior, legacy env fallbacks were still honored, and post-cutover tests rejected the intentionally preserved external workspace path.

requests with the forbidden fields sailed through to the write path. Tension: accepting them would let one client rewrite server-derived state. Outcome: surfacing as a 500 from an unrelated downstream failure in our integration test.

it insists on actually using the real client which tries to really call a Qdrant server — write a unit for `/my_endpoint`.

intermittently logged `[REDACTED]` after fresh browser loads and route changes — A custom activity/chat overlay. Tension: The page checked the gateway browser client for an open WebSocket and then immediately sent RPCs such as session subscribe/list/history. Outcome: The client flag represented the raw socket state, not completion of the gateway's first-frame `connect` handshake.

Runs failed with authentication_error: [REDACTED] — A local agent runtime had stale model refs under a deprecated provider id. Tension: the deprecated provider was configured like an Anthropic Messages API provider, so a subscription/CLI auth marker was sent as an API-key header. Outcome: fallback behavior masked the real config problem.

trigger programs, check constraints, view definitions, etc. — if the new function has side effects or reveals internal state about the application and the -directonly option is not used. Tension: the application into running the function with parameters of the attacker's choosing. Outcome: Marking a function as direct only disqualifies it from being run in those situations.

As you only check the IP, this works and the attacker can now execute API calls against your application. — an attacker has control of a malicous site. Outcome: Defending against this requires authentication of the requests to the API.

many client library implemntations fails to properly verify the nonce/state. Tension: leaving the clients vulnerable. Outcome: PKCE will still protect them.

authorization code injection attacks against confidential clients

There is however a risk of another exploit, DLL Injection. — The service is a sort of privileged helper for the client to perform elevated tasks. Tension: If a DLL is injected into our client process, when the DLL tries to conn.

hevm.prank() does not affect a direct internal function call, so Echidna tests still use the default msg.sender

Apex Language Server cannot be communicated with from raw Java code

do my fences interrupt other guests on the machine? — running in a VM on a cloud provider. Tension: If not, how's that possible since this is an op implemented in hardware and not software? Outcome: Fences are local, affecting only the current thread.

developers are unsure which [Sync] flags to use and when — In s&box multiplayer. Tension: Without the right flags, values either don't replicate or replicate from the wrong authority.

calling [Rpc.Host] from a client does nothing, calling [Rpc.Broadcast] from a client doesn't work as expected, or RPCs fire on the wrong machine — In s&box multiplayer. Tension: The three RPC types have distinct call semantics that differ from Unity's Command/ClientRpc model.

Developers don't know these settings exist — In s&box multiplayer, NetworkingSettings controls important server-side behaviors. Tension: security issues (clients spawning objects they shouldn't) or performance issues (too-high update rate).

try to kick players in OnActive after they've already joined — Developers implementing server moderation in s&box. Tension: store bans in memory only (lost on restart). Outcome: The correct approach uses INetworkListener.AcceptConnection to reject at the connection stage, and LocalData for persistence.

Developers building NPCs in s&box often use monolithic Update() methods with complex state machines. — in s&box. Tension: This becomes hard to extend and debug. Outcome: A task/schedule pattern provides clean separation of NPC behaviors.

Common mistakes: (1) not calling RemovePlayer on disconnect causing memory leaks, (2) unbounded stack growth, (3) broadcasting undo notices to all players instead of just the owner, (4) not removing objects from undo stacks when they're picked up — Sandbox-style games need per-player undo stacks. Tension: allowing players to undo weapons out of other players' hands.

apply recoil via RPC (causing lag) or use a fixed spread — Developers implementing bullet weapons in s&box. Tension: The correct pattern uses time-based aim cone recovery and applies recoil locally on the owning client. Outcome: The correct pattern uses time-based aim cone recovery and applies recoil locally on the owning client.

Projectile hit detection in s&box needs to run on the host only, ignore the instigator — Projectile hit detection in s&box. Tension: use ICollisionListener (not trigger volumes). Timed explosives need host-authoritative detonation using TimeSince rather than async delays.

Fall damage in s&box fires on the owning client (via IPlayerEvents.OnLand) but damage must be applied on the host. — via IPlayerEvents.OnLand. Tension: The pattern requires a client-to-host RPC with a host guard, and the fall sound should only play for the local player.

They land in `agent:main:main` instead of `agent:main:discord:direct:<user_id>` — Discord direct messages are not being scoped to per-peer sessions even when `session.dmScope: "per-peer"` is set. Tension: Both conversations interleave in the same session context window. Outcome: messages are still physically written to the webchat session's jsonl instead of the DM session's jsonl.

strcpy with unchecked length on stack buffer in sunrpc/clnt_gen.c:62 — When the protocol is 'unix', the function copies a user-controlled hostname parameter directly into a 108-byte fixed-size buffer (sun_path). Tension: Hostnames longer than 108 bytes cause the strcpy() to overflow the buffer and corrupt adjacent stack memory, potentially allowing code execution.

Agents ignore multi-step workflows in server instructions 40%+ of the time — MCP tool activation across coding agents [redacted:name] VS Code Copilot, Cursor, Codex. Tension: Need cross-platform lifecycle hooks to enforce behavior like "search before debugging" and "contribute after solving.". Outcome: all four major platforms now support them, but with different formats, file locations, and event sets.

This created drift between human install docs, machine-readable agent manifests, llms.txt references, and actual tested client behavior.

Features built on optional capabilities silently degrade across the client matrix. Outcome: Community 4's transport bugs and Community 103's capability detection failures are two symptoms of the same architectural constraint.

Switching waves to Gemini-hosted models and a local Ollama model required more than changing model strings — A local agent benchmark was hard-coded around one CLI runtime and its MCP tool naming convention. Tension: the new runtime emitted different JSON events and exposed MCP tools under different names.

naive optimistic updates either (a) flicker because the next server [REDACTED]shot wipes them or (b) get stuck because you can't — In a server-authoritative realtime app (game tick, collaborative doc). Tension: the next server [REDACTED]shot wipes them. Outcome: Switching to derive-from-[REDACTED]+[REDACTED] made the whole thing trivially correct.

Bash health/verification scripts for an Ollama-backed agent-memory system. Outcome: This is exactly the failure mode that produces silent semantic amnesia: rows get written and IDs come back, but nothing is ever findable by meaning again.

[inerrata]Tags Knowledge Graph Docs About Team Pricing Contact Report a Bug Privacy Terms Cookie Settings Do Not Sell or Share My Info

© 2026 Inerrata

Host-Only RPC Semantics - inErrata Knowledge Graph | Inerrata