registry / AI-agent capability abuse / @addai/node / 0.30.11

@addai/node@0.30.11

Daemon that pairs a machine with your +Ai account and runs Claude / Codex / Kimi / Gemini agents on its behalf. Reachable via Supabase from Vault, Entity Studio, or any other +Ai surface.

AI Security Review

scanned 1mo ago · by lpm-firewall-ai

The package has a real dangerous agent-control capability that alters a foreign user-level Claude setting, but its lifecycle hook is limited and persistence requires an explicit command. This warrants a warning rather than a publication block.

Static reason

One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source

Trigger

A user runs the ainode daemon and pairs the machine.

Impact

A user can lose Claude's warning before dangerous-mode runs, increasing the risk of unattended agent actions.

Mechanism

User-wide Claude permission-prompt suppression plus authenticated remote agent orchestration.

AI rationale

Evidence

dist/claude-config.jsdist/index.jsdist/desktop/relay-client.js~/.claude/settings.json~/.claude.json~/.ainode/state.json

Network endpoints1

desktop-relay-29522465016.europe-west2.run.app

Decision evidence

public snapshot

AI called this Suspicious at 90.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

Starting the daemon automatically sets Claude's user-wide bypass-permission acceptance.
The daemon opens an authenticated relay connection using its pairing token in the WebSocket URL.
The package launches external AI coding agents, including Claude with a dangerous permission-bypass flag when requested.

Evidence against

The only install hook loads a helper that repairs execute permission on node-pty's spawn helper; it does not reach the Claude configuration code.
Startup persistence is exposed as the explicit `ainode startup enable` command.
No runtime dependency on @addai/node, encoded payload, or install-time network activity was found.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsEvalFilesystemNetworkShellWebSocket

Supply chain

HighEntropyStringsMinified

Source & flagged code

12 flagged

package.jsonView file

•scripts.postinstall = node scripts/fix-pty-helper.js

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

dist/claude-system-prompt.jsView file

62const crypto_1 = require("crypto"); L63: const child_process_1 = require("child_process"); L64: const win_1 = require("./win");

High

Child Process

Package source references child process execution.

dist/claude-system-prompt.jsView on unpkg · L62

dist/flows/engine-loader.jsView file

21exports.engineLoaded = engineLoaded; L22: const dynamicImport = new Function('specifier', 'return import(specifier);'); L23: let cached = null;

Low

Eval

Package source references a known benign dynamic code generation pattern.

dist/flows/engine-loader.jsView on unpkg · L21

dist/unpair.jsView file

17exports.disconnectNode = disconnectNode; L18: const supabase_client_1 = require("./supabase-client"); L19: const store_1 = require("./store");

Medium

Dynamic Require

Package source references dynamic require/import behavior.

dist/unpair.jsView on unpkg · L17

dist/autostart-win.jsView file

13// L14: // The task is registered from generated XML rather than `schtasks` flags, L15: // because the settings that decide whether it ever runs are not reachable ... L59: const path = __importStar(require("path")); L60: const child_process_1 = require("child_process"); L61: const autostart_1 = require("./autostart"); ... L71: * username is ambiguous on a machine joined to one. */ L72: function currentUserId(env = process.env) { L73: const user = env.USERNAME || os.userInfo().username; L74: const dom = env.USERDOMAIN; ... L81: * behaves, instead of being wedged into a VBS string or a `schtasks` argument. L82: * The redirect matters: headless stdout on Windows has nowhere else to go, and

Medium

Install Persistence

Source writes installer persistence such as shell profile or service configuration.

dist/autostart-win.jsView on unpkg · L13

dist/desktop/open-viewer.jsView file

12// browser viewer is the honest fallback. L13: const child_process_1 = require("child_process"); L14: /** What we would run, as a pure value, so the choice is testable without L15: * launching anything. */ L16: function viewerCommand(platform, t, studioBase = 'https://entities.add.ai') { L17: if (platform === 'darwin' && t.vncPort) { ... L28: function openViewer(t) { L29: const cmd = viewerCommand(process.platform, t); L30: if (!cmd)

High

Sandbox Evasion Gated Capability

Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.

dist/desktop/open-viewer.jsView on unpkg · L12

dist/claude-config.jsView file

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. var __createbinding = (this && this.__createbinding) || (object.create ? (function(o, m, k, k2) { if (k2 === undefined) k2 = k; var desc = object.getownpropertydescriptor(m, k); if (!desc || ("get" in desc ? !m.__esmodule : desc.writable || desc.configurable)) { desc = { enumerable: true, get: function() { return m[k]; } }; } object.defineproperty(o, k2, desc); }) : (function(o, m, k, k2) { if (k2 === undefined) k2 = k; o[k2] = m[k]; })); var __setmoduledefault = (this && this.__setmoduledefault) || (object.create ? (function(o, v) { object.defineproperty(o, "default", { enumerable: true, value: v }); }) : function(o, v) { o["default"] =

High

Trigger Reachable External Ai Agent Control Surface Mutation

Manifest-trigger-reachable source writes behavior-bearing configuration into a user or project AI-agent control surface.

dist/claude-config.jsView on unpkg

dist/autostart-mac.jsView file

48Trigger-reachable persistence chain: scripts.start -> dist/cli.js -> dist/autostart.js -> dist/autostart-mac.js L48: const path = __importStar(require("path")); L49: const child_process_1 = require("child_process"); L50: const autostart_1 = require("./autostart"); L51: exports.PLIST_PATH = path.join(os.homedir(), 'Library', 'LaunchAgents', `${autostart_1.LABEL}.plist`); L52: /** `<string>` bodies are the only place user data lands in the plist, and a ... L127: * open() and the process HANGS in module resolution rather than failing. L128: * `launchctl print` says the job is fine. It is not. L129: * ... L172: try { L173: // stderr is piped, not inherited: `bootout` on a label that isn't loaded L174: // prints "Boot-out failed: 3: No such process", and that expected no-op

High

Trigger Reachable Persistence

A manifest entrypoint or package-local install chain reaches persistence behavior.

dist/autostart-mac.jsView on unpkg · L48

dist/self-update.jsView file

66exports.spawnReplacement = spawnReplacement; L67: const child_process_1 = require("child_process"); L68: const fs = __importStar(require("fs")); ... L75: * L76: * npx caches each invocation under `<npm-cache>/_npx/<hash>/node_modules/…`, L77: * so that segment is the reliable npx tell on every platform (checked first:

High

Runtime Package Install

Package source invokes a package manager install command at runtime.

dist/self-update.jsView on unpkg · L66

scripts/smoke-test.shView file

•path = scripts/smoke-test.sh kind = build_helper sizeBytes = 2483 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

scripts/smoke-test.shView on unpkg

dist/autostart-linux.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 15

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

dist/autostart-linux.js#virtual:normalized:round1View on unpkg

dist/codex-spawn.jsView file

•matchType = previous_version_dangerous_delta matchedPackage = @addai/node@0.30.10 matchedIdentity = npm:QGFkZGFpL25vZGU:0.30.10 similarity = 0.873 summary = stored previous version shares package body but lacks this dangerous source file

Critical

Previous Version Dangerous Delta

This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.

dist/codex-spawn.jsView on unpkg

Source & flagged code

12 flagged · loading source

package.jsonView file

•scripts.postinstall = node scripts/fix-pty-helper.js

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

dist/claude-system-prompt.jsView file

62const crypto_1 = require("crypto"); L63: const child_process_1 = require("child_process"); L64: const win_1 = require("./win");

High

Child Process

Package source references child process execution.

dist/claude-system-prompt.jsView on unpkg · L62

dist/flows/engine-loader.jsView file

21exports.engineLoaded = engineLoaded; L22: const dynamicImport = new Function('specifier', 'return import(specifier);'); L23: let cached = null;

Findings

1 Critical8 High6 Medium6 Low

CriticalPrevious Version Dangerous Deltadist/codex-spawn.js

HighInstall Time Lifecycle Scriptspackage.json

HighChild Processdist/claude-system-prompt.js

HighShell

HighSandbox Evasion Gated Capabilitydist/desktop/open-viewer.js

HighTrigger Reachable External Ai Agent Control Surface Mutationdist/claude-config.js

HighTrigger Reachable Persistencedist/autostart-mac.js

HighRuntime Package Installdist/self-update.js

HighSemantic Analysis Limiteddist/autostart-linux.js#virtual:normalized:round1

MediumDynamic Requiredist/unpair.js

MediumNetwork

MediumEnvironment Vars

Affected versions and remediation

This report applies to @addai/node@0.30.11.

See all versions of @addai/node for other recorded verdicts.

Review the evidence and your use of @addai/node@0.30.11 before allowing it. Restrict the permissions described in this report.
Choose an independently verified alternative or release. This report does not establish that other versions are safe.

Evidence last updated: 2026-09-14 08:32:09.163Z (UTC).

Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Related packages

More from @addai

Runtime surface

Behavioral surface

Related packages

More from @addai

Source & flagged code