Loading npm security reports…
Loading npm security reports…
@rubytech/create-realagent-code@0.1.458: AI-agent capability | LPM Firewall
Flagged as AI-agent capability risk
Allowed by default with warning: agent-facing configuration or capability changes need review before use.
AI security reviewscanned on Nov 8, 2026, 19:35 UTC · by lpm-firewall-ai The installer automatically marks the user's home workspace as trusted in Claude Code state, suppressing its workspace-trust prompt. The reviewed credential and network flows otherwise target their corresponding services, so this is a warning-level agent capability concern rather than evidence of credential theft.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review
Trigger
Running the package's create-realagent-code installer through its declared CLI entrypoint.
Impact
Claude Code may treat the user's home workspace as trusted without the user accepting its trust prompt, changing a broader agent setting during installation.
Mechanism
deployPayload() sets hasCompletedOnboarding and projects[$HOME].hasTrustDialogAccepted in the user's Claude state file.
Evidence
dist/index.js $HOME/<brand-config>/.claude/.claude.json
public snapshot
Evidence for warning
The package exposes dist/index.js as its command-line entrypoint and has no install lifecycle hook; its prepublishOnly script is for publishing. During payload deployment, the installer marks Claude Code onboarding complete and the home workspace trusted, then writes the state to the user's .claude.json file. The install sequence calls deployPayload(), which performs that state change. The reviewed credential flows send Claude OAuth refresh data to Claude's token endpoint, Anthropic API keys to api.anthropic.com, and Cloudflare tokens to api.cloudflare.com. Evidence against
The trust-state change is documented as a way to skip Claude Code's first-run prompts during installation, rather than as an unrelated hidden action. Reviewed network calls send credentials to the corresponding provider's own service; no foreign credential recipient was identified. Behavioral surface
Source Child process · Crypto · Dynamic require · Environment variables · Eval · Filesystem · Network · Shell
Supply chain High entropy strings · Minified · Possible obfuscation · Telemetry · URL strings
Manifest No license
Scanned 881 files · 12.7 MB of source
45 external domains referenced
0.0.0.0 127.0.0.1 api.anthropic.com api.cloudflare.com api.loop.software api.propertydata.co.uk api.replicate.com api.telegram.org appcenter.intuit.com astral.sh brew.sh chevrotain.io claude.ai deb.nodesource.com debian.neo4j.com dl.google.com docs.getmaxy.com docs.realagent.network en.wikipedia.org fonts.googleapis.com fonts.gstatic.com getmaxy.com github.com graph.microsoft.com langium.org llmstxt.org login.microsoftonline.com maxy.bot ns.adobe.com oauth.platform.intuit.com ollama.ai one.dash.cloudflare.com platform.claude.com player.vimeo.com quickbooks.api.intuit.com react.dev registry.npmjs.org sandbox-quickbooks.api.intuit.com schema.org www.anthropic.com www.apple.com www.linkedin.com www.w3.org www.xfa.org www.youtube-nocookie.com Findings4 critical · 22 high · 7 medium · 9 low Source & flagged code31 flagged Critical Download execute
Source downloads or fetches remote code and executes it.
dist/index.js View on unpkg · L1 High Sandbox evasion gated capability
Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.
dist/index.js View on unpkg · L1 Medium Install persistence
Source writes installer persistence such as shell profile or service configuration.
dist/index.js View on unpkg · L1 Shared code excerpt · 3 findings · 9 lines 1 #!/usr/bin/env node
2 import { execFileSync, spawn, spawnSync } from "node:child_process" ;
3 import { existsSync, mkdirSync, writeFileSync, cpSync, readFileSync, rmSync, readdirSync, appendFileSync, openSync, closeSync, chmodSync, statSync, realpathSync, accessSync, consta...
⋯
9 import { buildHeartbeatCronBlock, mergeHeartbeatBlock } from "./cron-registration.js" ;
10 import { validateTierFlag, validateEntitlementBase64, applyTierToAccountConfig, entitlementPath } from "./tier-flag.js" ;
11 import { parseOsRelease, isUbuntuLike as isUbuntuLikePure, parseAptCacheCandidate, decideAptResolution, } from "./apt-resolve.js" ;
⋯
51 }
52 const INSTALL_DIR = resolve (process.env.HOME ?? "/root" , BRAND.installDir);
53 const PERSIST_DIR = resolve (process.env.HOME ?? "/root" , BRAN
Critical Trigger reachable dangerous capability
A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.
dist/index.js View on unpkg · L1 Code excerpt · 9 lines Trigger-reachable chain: manifest.bin -> dist/index.js
1 #!/usr/bin/env node
2 import { execFileSync, spawn, spawnSync } from "node:child_process" ;
3 import { existsSync, mkdirSync, writeFileSync, cpSync, readFileSync, rmSync, readdirSync, appendFileSync, openSync, closeSync, chmodSync, statSync, realpathSync, accessSync, consta...
⋯
9 import { buildHeartbeatCronBlock, mergeHeartbeatBlock } from "./cron-registration.js" ;
10 import { validateTierFlag, validateEntitlementBase64, applyTierToAccountConfig, entitlementPath } from "./tier-flag.js" ;
11 import { parseOsRelease, isUbuntuLike as isUbuntuLikePure, parseAptCacheCandidate, decideAptResolution, } from "./apt-resolve.js" ;
⋯
51 }
52
High Child process
Package source references child process execution.
dist/index.js View on unpkg · L1 Code excerpt · 3 lines 1 #!/usr/bin/env node
2 import { execFileSync, spawn, spawnSync } from "node:child_process" ;
3 import { existsSync, mkdirSync, writeFileSync, cpSync, readFileSync, rmSync, readdirSync, appendFileSync, openSync, closeSync, chmodSync, statSync, realpathSync, accessSync, consta...
Code excerpt · 3 lines 536 // Fetch + dearmor the signing key, write to /etc/apt/trusted.gpg.d/. Pipe
537 // composition runs through bash -c so the curl|gpg pipeline is one
538 // privileged command rather than two separate sudo escalations.
High Trigger reachable persistence
A manifest entrypoint or package-local install chain reaches persistence behavior.
dist/index.js View on unpkg · L1 Code excerpt · 8 lines Trigger-reachable persistence chain: manifest.bin -> dist/index.js
1 #!/usr/bin/env node
2 import { execFileSync, spawn, spawnSync } from "node:child_process" ;
3 import { existsSync, mkdirSync, writeFileSync, cpSync, readFileSync, rmSync, readdirSync, appendFileSync, openSync, closeSync, chmodSync, statSync, realpathSync, accessSync, consta...
⋯
9 import { buildHeartbeatCronBlock, mergeHeartbeatBlock } from "./cron-registration.js" ;
10 import { validateTierFlag, validateEntitlementBase64, applyTierToAccountConfig, entitlementPath } from "./tier-flag.js" ;
11 import { parseOsRelease, isUbuntuLike as isUbuntuLikePure, parseAptCacheCandidate, decideAptResolution, } from "./apt-resolve.js" ;
⋯
51 }
52 INSTALL_DIR = (process.env.HOME ?? , BRAND.installDir);
High Runtime package install
Package source invokes a package manager install command at runtime.
dist/index.js View on unpkg · L284 Code excerpt · 6 lines 284 function canSudo () {
285 const result = spawnSync ("sudo" , ["-n" , "true" ], { stdio: "pipe" , timeout: 5_000 });
286 return result.status === 0 ;
⋯
288 /** True when the npm global prefix's node_modules dir is writable by the current
289 * user (nvm/Homebrew installs) — meaning `npm install -g` needs no sudo. */
290 function npm[redacted]() {
High Known malware source similarity
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
dist/index.js View on unpkg Scanner record · 6 lines matchType = normalized_sha256
matchedPackage = @rubytech/create-sitedesk-code@0.1.459
matchedPath = dist/index.js
matchedIdentity = npm:[redacted]:0.1.459
similarity = 1.000
summary = normalized source hash matched finalized malicious source
Critical Credential exfiltration
Source appears to send environment or credential material to an external endpoint.
payload/server/server.js View on unpkg · L358 High Command output exfiltration
Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.
payload/server/server.js View on unpkg · L358 Shared code excerpt · 2 findings · 12 lines 358 });
359 process.stderr.write (`[trash:marked] accountId=${accountId} elementId=${elementId} labels=${baseLabels.join(",")} by=${by} reason=${reason ?? "null"}
360 `);
⋯
383 const keysJson = lookup.records[0 ].get ("keysJson" );
384 const originalKeys = keysJson ? JSON.parse (keysJson) : {};
High Known malware source similarity
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
dist/uninstall.js View on unpkg Scanner record · 6 lines matchType = normalized_sha256
matchedPackage = @rubytech/create-sitedesk-code@0.1.459
matchedPath = dist/uninstall.js
matchedIdentity = npm:[redacted]:0.1.459
similarity = 1.000
summary = normalized source hash matched finalized malicious source
Code↓
Package metadata 11 findings
Medium Network
Medium Environment variables
Medium Structural risk force deep review
Low Non install lifecycle scripts
const INSTALL_DIR = resolve (process.env.HOME ?? "/root" , BRAND.installDir);
const
resolve
"/root"
11
* 2 . Installs the Maxy stderr tee so upstream stderr lands in the
12 * per-conversation stream log alongside every other plugin.
26 import { execFileSync, spawn } from "node:child_process" ;
27 import { accessSync, appendFileSync, constants, mkdirSync, readFileSync, statSync } from "node:fs" ;
76 const iso = new Date ().toISOString ();
77 const logDir = process.env.LOG_DIR;
96 process.stderr.write (`${msg}\n` );
⋯
10 const platformRoot = process.env.PLATFORM_ROOT;
High Known malware source similarity
Source file is highly similar to a previously finalized malicious package; route for source-aware review.
payload/platform/plugins/cloudflare/mcp/dist/lib/cloudflared.js View on unpkg Scanner record · 6 lines matchType = normalized_sha256
matchedPackage = @rubytech/create-realagent-code@0.1.392
matchedPath = [redacted].js
matchedIdentity = npm:[redacted]:0.1.392
similarity = 1.000
summary = normalized source hash matched finalized malicious source
const
new
Promise
14 (0 , node_child_process_1.execFile)(cmd, args, { env }, (err, stdout, stderr) => {
24 return JSON.parse (raw);
385 for (const label of baseLabels) {
745 const colour = colourForLabel2 (label);
746 const svg = "<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 24 24'><circle cx='12' cy='12' r='12' fill='" + colour + "'/><g transform='translate(4.8 4.8) scale(0.6)' fill='no...
747 return "data:image/svg+xml," + encodeURIComponent (svg);
1187 const accountId = props.accountId;
1188 const expectedAccountId = params.expectedAccountId ?? process.env.ACCOUNT_ID;
from
"node:fs"
19 import { createHash, randomInt, randomUUID } from "node:crypto" ;
20 import { createConnection } from "node:net" ;
21 import { homedir, hostname as osHostname } from "node:os" ;
50 const PLATFORM_ROOT = process.env.PLATFORM_ROOT ?? resolve (import .meta.dirname, "../../
path = payload/platform/plugins/prompt-optimiser/skills/prompt-optimiser/__tests__/restatement-ladder.test.sh