@m8t-stack/cli@0.2.164

CLI for managing m8t deployments on Azure / Microsoft Agent Foundry

AI Security Review

scanned 23d ago · by lpm-firewall-ai

The inspected source shows package-aligned deployment and companion operations without a concrete malicious chain. The flagged agent integration is a read-only drift check, and gateway credentials are freshly generated rather than copied from an existing service session.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

Users invoke CLI commands; npm installation has no automatic install lifecycle hook.

Impact

No malicious agent control mutation or credential exfiltration was established.

Mechanism

The agent integration reads skill sidecars, while gateway provisioning uses a newly generated bearer and authenticates Azure requests to Azure.

AI rationale

Network endpoints1

management.azure.com

Decision evidence

public snapshot

AI called this Clean at 88.0% confidence as Benign with low false-positive risk.

Evidence for AI clean decision

package.json defines no registry install lifecycle hooks or runtime self-dependency.
The indicated agent integration locates m8t-architect skill metadata and reads it for a drift check; it does not modify agent configuration.
Gateway setup generates a fresh bearer rather than forwarding an existing service session credential.
Azure authentication is sent to management.azure.com when provisioning the gateway connection.
Companion installation runs through a CLI command and checks package-owned destination boundaries.

Evidence against

Behavioral surface

Source

ChildProcessCryptoEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStringsUrlStrings

ManifestNo manifest risk signals triggered.

scanned 1 file(s), 1.51 MB of source, external domains: acc.services.ai.azure.com, ai.azure.com, aka.ms, api.github.com, api.githubcopilot.com, api.openai.com, auth.openai.com, companion.invalid, docs.expo.dev, gateway.example.com, ghcr.io, github.com, graph.microsoft.com, help.openai.com, learn.microsoft.com, m8t-admin.wonderfuldesert-721e332f.eastus2.azurecontainerapps.io, m8t-kv.vault.azure.net, management.azure.com, myaccount.services.ai.azure.com, portal.azure.com, raw.githubusercontent.com, vault.azure.net, www.apple.com

Source & flagged code

7 flagged

dist/cli.jsView file

130Source sends the broad process environment to a literal external destination. L130: try { L131: const obj2 = JSON.parse(raw); L132: if (!obj2 || typeof obj2.repo !== "string" || obj2.repo.length === 0) ... L343: // ../../packages/github-app-auth/dist/esm/jwt.js L344: import { createSign, createPrivateKey } from "crypto"; L345: function signAppJwt(args) { ... L356: sign.update(unsigned); L357: sign.end(); L358: const signature = sign.sign(createPrivateKey(args.privateKeyPem)).toString("base64url"); ... L461: const appJwt = signAppJwt({ appId, privateKeyPem }); L462: const url = `https://api.github.com/app/installations/${args.installationId}/access_tokens`; L463: const repoName = args.repository.includes("/") ? args.repository.split("/", 2)[1] : args.repository;

Critical

Hardcoded Runtime Data Exfiltration

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

dist/cli.jsView on unpkg · L130

12936init_errors(); L12937: import { spawn } from "child_process"; L12938: var SECRET_VALUE_FLAGS = /* @__PURE__ */ new Set(["--secure-environment-variables"]);

High

Child Process

Package source references child process execution.

dist/cli.jsView on unpkg · L12936

•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.js: Authorization: `Bearer ${appJwt}`, Authorization: `Bearer ${appJwt}`, Authorization: `Bearer ${appJwt}`, Authorization: `Bearer ${token.token}`, headers: { Authorization: `Bearer ${token.token}` } headers: { Authorization: `Bearer ${token.token}`, "Content-Type": "application/json", ...args.extraHeaders ?? {} }, headers: { Authorization: `Bearer ${token.token}`, "Content-Type": "application/json", ...args.extraHeaders ?? {} }, body: JSON.stringify({ agent_endpoint: { authorization_schemes: [{ type: "Entra", isolation_key_source: { kind: "Entra" } }] } })

High

Credential Redirect Persistence

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

dist/cli.jsView on unpkg

29246); L29247: this.context.stdout.write( L29248: colors.dim("next: run 'm8t doctor' to verify gateway + data-plane access.\n") ... L29254: // src/commands/open.ts L29255: import { spawn as spawn5 } from "child_process"; L29256: import { Command as Command55, Option as Option52 } from "clipanion"; ... L29260: var OPEN_TARGETS = ["webapp", "foundry", "portal"]; L29261: var FOUNDRY_PORTAL_URL = "https://ai.azure.com"; L29262: function azurePortalRgUrl(tenantId, subscriptionId, resourceGroup) {

High

Command Output Exfiltration

Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.

dist/cli.jsView on unpkg · L29246

130try { L131: const obj2 = JSON.parse(raw); L132: if (!obj2 || typeof obj2.repo !== "string" || obj2.repo.length === 0) ... L343: // ../../packages/github-app-auth/dist/esm/jwt.js L344: import { createSign, createPrivateKey } from "crypto"; L345: function signAppJwt(args) { ... L356: sign.update(unsigned); L357: sign.end(); L358: const signature = sign.sign(createPrivateKey(args.privateKeyPem)).toString("base64url"); ... L461: const appJwt = signAppJwt({ appId, privateKeyPem }); L462: const url = `https://api.github.com/app/installations/${args.installationId}/access_tokens`; L463: const repoName = args.repository.includes("/") ? args.repository.split("/", 2)[1] : args.repository;

High

Sandbox Evasion Gated Capability

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

dist/cli.jsView on unpkg · L130

2163} L2164: throw new Error("`fetch` is not defined as a global; Either pass `fetch` to the client, `new OpenAI({ fetch })` or polyfill the global, `globalThis.fetch = fetch`"); L2165: } ... L12936: init_errors(); L12937: import { spawn } from "child_process"; L12938: var SECRET_VALUE_FLAGS = /* @__PURE__ */ new Set(["--secure-environment-variables"]); ... L14647: try { L14648: creds = JSON.parse(fs3.readFileSync(credsPath, "utf8")); L14649: } catch { ... L37156: args.stdout( L37157: `${colors.success("\u2705 The platform is live and your local tools are pointed at it.")} L37158: gateway: ${d.gatewayUrl}

High

Remote Agent Bridge

Source exposes local file and command tools to a remote model endpoint.

dist/cli.jsView on unpkg · L2163

29246Trigger-reachable command-output exfiltration chain: manifest.main -> dist/cli.js L29246: ); L29247: this.context.stdout.write( L29248: colors.dim("next: run 'm8t doctor' to verify gateway + data-plane access.\n") ... L29254: // src/commands/open.ts L29255: import { spawn as spawn5 } from "child_process"; L29256: import { Command as Command55, Option as Option52 } from "clipanion"; ... L29260: var OPEN_TARGETS = ["webapp", "foundry", "portal"]; L29261: var FOUNDRY_PORTAL_URL = "https://ai.azure.com"; L29262: function azurePortalRgUrl(tenantId, subscriptionId, resourceGroup) {

High

Trigger Reachable Command Output Exfiltration

A manifest entrypoint or package-local install chain reaches command-output exfiltration behavior.

dist/cli.jsView on unpkg · L29246

Source & flagged code

7 flagged · loading source

dist/cli.jsView file

Critical

Hardcoded Runtime Data Exfiltration

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

dist/cli.jsView on unpkg · L130

12936init_errors(); L12937: import { spawn } from "child_process"; L12938: var SECRET_VALUE_FLAGS = /* @__PURE__ */ new Set(["--secure-environment-variables"]);

High

Child Process

Package source references child process execution.

dist/cli.jsView on unpkg · L12936

Findings

1 Critical7 High3 Medium5 Low

CriticalHardcoded Runtime Data Exfiltrationdist/cli.js

HighChild Processdist/cli.js

HighShell

HighCredential Redirect Persistencedist/cli.js

HighCommand Output Exfiltrationdist/cli.js

HighSandbox Evasion Gated Capabilitydist/cli.js

HighRemote Agent Bridgedist/cli.js

HighTrigger Reachable Command Output Exfiltrationdist/cli.js

MediumNetwork

MediumEnvironment Vars

MediumStructural Risk Force Deep Review

LowNon Install Lifecycle Scripts

LowScripts Present

LowFilesystem

Affected versions and remediation

This report applies to @m8t-stack/cli@0.2.164.

See all versions of @m8t-stack/cli for other recorded verdicts.

Review the evidence and your use of @m8t-stack/cli@0.2.164 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: 2026-09-30 04:49:37.938Z (UTC).

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