registry / AI-agent capability abuse / create-mastra / 1.27.2

create-mastra@1.27.2

Create Mastra apps with one command

AI Security Review

scanned 2mo ago · by lpm-firewall-ai

The source shows an explicit-user-command agent setup path and default telemetry, which warrant a warning. It does not show automatic npm lifecycle execution or a concrete malicious payload.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

The user runs the create-mastra command without --no-skills.

Impact

A remote installer may modify supported AI-agent integrations in the generated project; telemetry exposes host and CLI metadata.

Mechanism

Remote skills installation and default CLI telemetry.

AI rationale

The source shows an explicit-user-command agent setup path and default telemetry, which warrant a warning. It does not show automatic npm lifecycle execution or a concrete malicious payload.

Evidence

package.jsondist/index.js.mastra/analytics.json<generated-project-path>

Network endpoints1

us.posthog.com

Decision evidence

public snapshot

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

Evidence for warning

The user-invoked create command installs Mastra skills through npx for detected coding agents by default.
The CLI enables PostHog telemetry by default and sends a persistent identifier, hostname, and system details.
The skills setup is remote command execution outside the package and can affect detected AI-agent integrations.

Evidence against

package.json has no preinstall, install, or postinstall hook.
Project generation, dependency installation, and skills setup occur only after the user runs the create command.
Platform observability requires an interactive choice before its authentication flow.

Behavioral surface

Source

ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStringsTelemetryUrlStrings

ManifestNo manifest risk signals triggered.

scanned 2 file(s), 496 KB of source, external domains: app.posthog.com, gateway-api.mastra.ai, gateway-api.staging.mastra.ai, github.com, mastra.ai, platform.mastra.ai, projects.mastra.ai, qntm.org, raw.githubusercontent.com, studio.mastra.ai, studio.staging.mastra.ai, us.posthog.com

Source & flagged code

10 flagged

dist/index.jsView file

Lines 1-31javascript

1#! /usr/bin/env node

2import { InvalidArgumentError, Command } from 'commander';

3import { randomBytes, randomUUID } from 'node:crypto';

4import fs, { readFileSync, statSync, appendFileSync, writeFileSync, createWriteStream, createReadStream, existsSync, mkdirSync, constants as constants$2 } from 'node:fs';

5import os, { homedir, release, constants as constants$1 } from 'node:os';

6import path, { join, dirname } from 'node:path';

7import * as fsPromises from 'node:fs/promises';

8import fsPromises__default, { readFile, mkdir, writeFile, chmod } from 'node:fs/promises';

9import require$$0, { execFileSync, ChildProcess, execFile, spawnSync, spawn } from 'node:child_process';

10import { createServer } from 'node:http';

High

Child Process

Package source references child process execution.

dist/index.jsView on unpkg · L8

11import require$$4, { styleText, debuglog, stripVTControlCharacters, inspect, promisify, callbackify, aborted } from 'node:util';

12import process$2, { stdout, stdin, platform, hrtime, execPath, execArgv } from 'node:process';

13import * as l from 'node:readline';

Critical

Hardcoded Runtime Data Exfiltration

Source sends credentials or rich application records to a package-controlled external receiver enabled by default.

dist/index.jsView on unpkg · L8

Critical

Trigger Reachable Dangerous Capability

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

dist/index.jsView on unpkg · L8

High

Shell

Package source references shell execution.

dist/index.jsView on unpkg · L11

High

Sandbox Evasion Gated Capability

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

dist/index.jsView on unpkg · L8

14import l__default from 'node:readline';

15import tty, { ReadStream } from 'node:tty';

16import { PostHog } from 'posthog-node';

17import { fileURLToPath } from 'node:url';

18import { StringDecoder } from 'node:string_decoder';

19import { setTimeout as setTimeout$1, scheduler, setImmediate } from 'node:timers/promises';

20import { once, addAbortListener, EventEmitter, on, setMaxListeners } from 'node:events';

21import { serialize } from 'node:v8';

22import { Transform, getDefaultHighWaterMark, Duplex, Writable, Readable, PassThrough } from 'node:stream';

23import { Buffer as Buffer$1 } from 'node:buffer';

24import { finished } from 'node:stream/promises';

25import { x } from 'tinyexec';

27const PATH_PARAM_RE = /\{[^{}]+\}/g;

28const supportsRequestInitExt = () => {

29 return typeof process === "object" && Number.parseInt(process?.versions?.node?.substring(0, 2)) >= 18 && process.versions.undici;

30};

31function randomID() {

Lines 10317-10357javascript

10317// Wraps `finished(stream)` to handle the following case:

10318// - When the subprocess exits, Node.js automatically calls `subprocess.stdin.destroy()`, which we need to ignore.

10319// - However, we still need to throw if `subprocess.stdin.destroy()` is called before subprocess exit.

10320const waitForStream = async (stream, fdNumber, streamInfo, {isSameDirection, stopOnExit = false

Lines 15209-15249javascript

15209 });

15210 } catch (gitError) {

15211 if (signal?.aborted) signal.throwIfAborted();

15212 throw new Error(`Failed to clone repository: ${gitError instanceof Error ? gitError.

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. dist/index.js: import fs, { readFileSync, statSync, appendFileSync, writeFileSync, createWriteStream, createReadStream, existsSync, mkdirSync, constants as constants$2 } from 'node:fs'; import fsPromises__default, { readFile, mkdir, writeFile, chmod } from 'node:fs/promises'; import * as l from 'node:readline'; import l__default from 'node:readline'; baseUrl = "", baseUrl = removeTrailingSlash(baseUrl); baseUrl: localBaseUrl, let finalBaseUrl = baseUrl;

High

Credential Redirect Persistence

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

dist/index.jsView on unpkg

•Manifest-reachable source resolves another installed package, overwrites its runtime code, and injects package-defined remote behavior. dist/index.js: import fs, { readFileSync, statSync, appendFileSync, writeFileSync, createWriteStream, createReadStream, existsSync, mkdirSync, constants as constants$2 } from 'node:fs'; import fsPromises__default, { readFile, mkdir, writeFile, chmod } from 'node:fs/promises'; redirect: "follow", const MASTRA_PLATFORM_API_URL = process.env.MASTRA_PLATFORM_API_URL || "https://platform.mastra.ai"; if (MASTRA_PLATFORM_API_URL.includes("staging")) return "https://gateway-api.staging.mastra.ai/v1"; return "https://gateway-api.mastra.ai/v1"; if (MASTRA_PLATFORM_API_URL.includes("staging")) return "https://studio.staging.mastra.ai"; return "https://studio.mastra.a

High

Entrypoint Foreign Package Code Overwrite

Manifest-reachable source overwrites another installed package with package-defined remote behavior.

dist/index.jsView on unpkg

Source & flagged code

10 flagged · loading source

dist/index.jsView file

8import fsPromises__default, { readFile, mkdir, writeFile, chmod } from 'node:fs/promises'; L9: import require$$0, { execFileSync, ChildProcess, execFile, spawnSync, spawn } from 'node:child_process'; L10: import { createServer } from 'node:http'; L11: import require$$4, { styleText, debuglog, stripVTControlCharacters, inspect, promisify, callbackify, aborted } from 'node:util'; L12: import process$2, { stdout, stdin, platform, hrtime, execPath, execArgv } from 'node:process'; L13: import * as l from 'node:readline'; ... L86: // with no body, we should not to set Content-Type L87: serializedBody === void 0 || // if serialized body is FormData; browser will correctly set Content-Type & boundary expression L88: serializedBody instanceof FormData ? {} : { ... L214: if (parseAs === "json" &

Critical

Hardcoded Runtime Data Exfiltration

Source sends credentials or rich application records to a package-controlled external receiver enabled by default.

dist/index.jsView on unpkg · L8

8Trigger-reachable chain: manifest.main -> dist/index.js L8: import fsPromises__default, { readFile, mkdir, writeFile, chmod } from 'node:fs/promises'; L9: import require$$0, { execFileSync, ChildProcess, execFile, spawnSync, spawn } from 'node:child_process'; L10: import { createServer } from 'node:http'; L11: import require$$4, { styleText, debuglog, stripVTControlCharacters, inspect, promisify, callbackify, aborted } from 'node:util'; L12: import process$2, { stdout, stdin, platform, hrtime, execPath, execArgv } from 'node:process'; L13: import * as l from 'node:readline'; ... L86: // with no body, we should not to set Content-Type L87: serializedBody === void 0 || // if serialized body is FormData; browser will correctly set Content-Type & boundary expression L88: serializedBody instanc

Findings

2 Critical8 High3 Medium5 Low

CriticalHardcoded Runtime Data Exfiltrationdist/index.js

CriticalTrigger Reachable Dangerous Capabilitydist/index.js

HighChild Processdist/index.js

HighShelldist/index.js

HighSame File Env Network Executiondist/index.js

HighCredential Redirect Persistencedist/index.js

HighEntrypoint Foreign Package Code Overwritedist/index.js

HighCommand Output Exfiltrationdist/index.js

HighSandbox Evasion Gated Capabilitydist/index.js

HighTrigger Reachable Command Output Exfiltrationdist/index.js

MediumNetwork

MediumEnvironment Vars

Affected versions and remediation

This report applies to create-mastra@1.27.2.

See all versions of create-mastra for other recorded verdicts.

Review the evidence and your use of create-mastra@1.27.2 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: 2026-08-28 21:39:37.325Z (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

Runtime surface

Behavioral surface

Source & flagged code