registry / @agentgates/cli / 0.15.0

@agentgates/cli@0.15.0

Wallet operations for AI agents and humans

AI Security Review

scanned 3h ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Installation alters AI-agent instruction and rule files in the user's home directory. It injects content directing agents to prefer this package for wallet operations.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm installation invokes postinstall automatically.

Impact

Can steer installed AI agents toward package-controlled wallet commands without user consent.

Mechanism

Automatic injection into Claude Code, Cursor, and Windsurf control surfaces.

Policy narrative

On installation, package.json runs scripts/postinstall.mjs. That script detects AI coding tools and writes an instruction block into Claude Code, Cursor, and Windsurf configuration files. The block directs AI agents to use the package CLI for wallet operations, changing external agent behavior without an explicit user command.

Rationale

This is an unconsented postinstall mutation of multiple foreign AI-agent control surfaces. The automatic lifecycle trigger and behavior-steering content meet the blocking policy.

Evidence

package.jsonscripts/postinstall.mjs~/.claude/CLAUDE.md~/.cursor/rules/agentgates.md~/.windsurfrules

Decision evidence

public snapshot

AI identified this as a dangerous AI-agent capability at 98.0% confidence with low false-positive risk.

Evidence for policy risk

The install lifecycle automatically runs the postinstall script.
The postinstall script injects instructions into the user's Claude Code control file.
The postinstall script creates and writes a Cursor rule file without user action.

Evidence against

The reviewed install hook does not contact a network endpoint.
No runtime self-dependency is declared.

Behavioral surface

Source

ChildProcessEnvironmentVarsFilesystemNativeBindingsNetwork

Supply chain

HighEntropyStringsPossibleObfuscationUrlStrings

ManifestNo manifest risk signals triggered.

scanned 1 file(s), 261 KB of source, external domains: api.devnet.solana.com

Oversized source lightweight scan

dist/cli.mjs2.75 MB file, sampled 256 KB

FilesystemNetworkChildProcessEnvironmentVarsNativeBindingsPossibleObfuscationHighEntropyStringsUrlStringsapi.devnet.solana.com

Source & flagged code

10 flagged

package.jsonView file

Lines 17-57json

17 "license": "MIT",

18 "homepage": "https://agentgates.ai",

19 "repository": {

20 "type": "git",

21 "url": "git+https://github.com/Carlos-Sotop/agentgates-ai-cli.git"

22 },

23 "author": "Agentgates",

24 "type": "module",

25 "bin": {

26 "agentgates": "bin/agentgates"

27 },

28 "files": [

29 "bin",

30 "dist/cli.mjs",

31 "scripts/postinstall.mjs",

32 "README.md"

33 ],

34 "scripts": {

35 "build": "esbuild src/main.ts --bundle --platform=node --format=esm --outfile=dist/cli.mjs --banner:js=\"import{createRequire}from'module';const require=createRequire(import.meta.url);\"",

36 "build:tsc": "tsc -p tsconfig.json",

37 "postinstall": "node scripts/postinstall.mjs || exit 0",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L37

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg · L37

38 "prepare": "[ -x node_modules/.bin/esbuild ] && npm run build || true",

39 "dev": "tsc -w -p tsconfig.json",

40 "typecheck": "tsc --noEmit",

41 "bounty": "node bin/bounty",

42 "sync:bounty-shared": "node scripts/sync-bounty-shared.mjs",

43 "clean": "rm -rf dist"

44 },

45 "engines": {

46 "node": ">=20"

47 },

48 "devDependencies": {

49 "@solana/web3.js": "^1.98.4",

50 "@types/node": "^22.10.0",

51 "commander": "^12.1.0",

52 "esbuild": "^0.28.0",

53 "ethers": "^6.16.0",

54 "open": "^10.1.0",

55 "picocolors": "^1.1.1",

56 "typescript": "^5.6.3"

57 }

dist/cli.mjsView file

56* Constructs the CommanderError class L57: * @param {number} exitCode suggested exit code which could be used with process.exit L58: * @param {string} code an id string representing the error ... L982: var EventEmitter = __require("node:events").EventEmitter; L983: var childProcess2 = __require("node:child_process"); L984: var path6 = __require("node:path"); ... L1031: this._outputConfiguration = { L1032: writeOut: (str) => process8.stdout.write(str), L1033: writeErr: (str) => process8.stderr.write(str), ... L1067: * @returns {Command[]} L1068: * @private L1069: */

Critical

Wallet Drain

Source uses private key material to transfer cryptocurrency funds.

dist/cli.mjsView on unpkg · L56

56Trigger-reachable chain: manifest.bin -> bin/agentgates -> dist/cli.mjs L56: * Constructs the CommanderError class L57: * @param {number} exitCode suggested exit code which could be used with process.exit L58: * @param {string} code an id string representing the error ... L982: var EventEmitter = __require("node:events").EventEmitter; L983: var childProcess2 = __require("node:child_process"); L984: var path6 = __require("node:path"); ... L1031: this._outputConfiguration = { L1032: writeOut: (str) => process8.stdout.write(str), L1033: writeErr: (str) => process8.stderr.write(str), ... L1067: * @returns {Command[]} L1068: * @private L1069: */

Critical

Trigger Reachable Dangerous Capability

A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.

dist/cli.mjsView on unpkg · L56

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. dist/cli.mjs: delete opts.headers.authorization; if (opts.auth && !options.headers.authorization) { options.headers.authorization = "Basic " + Buffer.from(opts.auth).toString("base64"); return url.host === null || url.host === "" || url.cannotBeABaseURL || url.scheme === "file"; cannotBeABaseURL: false this.url.cannotBeABaseURL = true; if (this.base === null || this.base.cannotBeABaseURL && c !== 35) { } else if (this.base.cannotBeABaseURL && c === 35) {

scripts/postinstall.mjsView file

Lines 1-21javascript

1#!/usr/bin/env node

2import { existsSync, readFileSync, writeFileSync, mkdirSync } from "fs";

3import { join } from "path";

Critical

Ai Agent Control Hijack

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

scripts/postinstall.mjsView on unpkg · L1

4import { homedir } from

Source & flagged code

10 flagged · loading source

package.jsonView file

•scripts.postinstall = node scripts/postinstall.mjs || exit 0

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

•scripts.postinstall = node scripts/postinstall.mjs || exit 0

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg

dist/cli.mjsView file

Critical

Wallet Drain

Source uses private key material to transfer cryptocurrency funds.

dist/cli.mjsView on unpkg · L56

Critical

Trigger Reachable Dangerous Capability

A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.

dist/cli.mjsView on unpkg · L56

High

Credential Redirect Persistence

Manifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.

dist/cli.mjsView on unpkg

27008var _256n2 = BigInt(256); L27009: var _0x71n2 = BigInt(113); L27010: var SHA3_PI2 = [];

High

Obfuscated Payload Loader

Source contains an obfuscated payload loader that reconstructs and executes hidden code.

dist/cli.mjsView on unpkg · L27008

•stage = ast_semantic_analysis; reason = ast_alias_growth_limit_exceeded; limitedFiles = 1

High

Semantic Analysis Limited

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

dist/cli.mjsView on unpkg

Low

Weak Crypto

Package source references weak cryptographic algorithms.

dist/cli.mjsView on unpkg · L56

scripts/postinstall.mjsView file

1#!/usr/bin/env node L2: import { existsSync, readFileSync, writeFileSync, mkdirSync } from "fs"; L3: import { join } from "path"; ... L48: if (!existsSync(filePath)) { L49: writeFileSync(filePath, wrapped + "\n", "utf-8"); L50: return "created"; ... L58: ); L59: writeFileSync(filePath, content.replace(re, wrapped), "utf-8"); L60: return "updated"; ... L62: L63: writeFileSync(filePath, content.trimEnd() + "\n\n" + wrapped + "\n", "utf-8"); L64: return "appended";

Critical

Ai Agent Control Hijack

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

scripts/postinstall.mjsView on unpkg · L1

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. ns the lane itself over signed http. \`\`\` agentgates passwords init # mint twelve words + the key (once); --seed "<words>" brings words; agentgates_passwords_seed works too agentgates passwords add --label <name> --host <site> --username <u> --password <p> agentgates passwords allow [--agent <addr>] [--ask-me] # permit yourself, or another agent agentgates passwords use --url https://<site>/login # log in through the box, get the page back agentgates passwords list # what this key holds and may use agentgates passwords invite # print a link a person opens to take the ow

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.

scripts/postinstall.mjsView on unpkg

Findings

3 Critical5 High4 Medium6 Low

CriticalWallet Draindist/cli.mjs

CriticalAi Agent Control Hijackscripts/postinstall.mjs

CriticalTrigger Reachable Dangerous Capabilitydist/cli.mjs

HighInstall Time Lifecycle Scriptspackage.json

HighCredential Redirect Persistencedist/cli.mjs

HighObfuscated Payload Loaderdist/cli.mjs

HighTrigger Reachable External Ai Agent Control Surface Mutationscripts/postinstall.mjs

HighSemantic Analysis Limiteddist/cli.mjs

MediumAmbiguous Install Lifecycle Scriptpackage.json

MediumNetwork

MediumEnvironment Vars

MediumStructural Risk Force Deep Review

LowNon Install Lifecycle Scripts

LowScripts Present

LowWeak Cryptodist/cli.mjs

LowFilesystem

LowHigh Entropy Strings

LowUrl Strings