registry / AI-agent capability abuse / alink-cli / 0.11.14

alink-cli@0.11.14

一条命令把工作机接入 AgentLink，随时随地遥控本机的编码 agent。One command to link your machine to AgentLink and control your coding agents from anywhere.

AI Security Review

scanned 2mo ago · by lpm-firewall-ai

Source inspection confirms a user-invoked remote coding-agent control capability with optional approvals and session access. It is not concrete malware, but its capability warrants a warning.

Static reason

High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review

Trigger

The user runs alink-cli and pairs or supplies a hub credential.

Impact

A party controlling the paired hub or credential can direct local coding agents within allowed directories; agent actions may affect local projects.

Mechanism

WebSocket-controlled coding-agent execution and session access.

AI rationale

Source inspection confirms a user-invoked remote coding-agent control capability with optional approvals and session access. It is not concrete malware, but its capability warrants a warning.

Evidence

package.jsonREADME.mdbin/agentlink.jsdist/daemon.js.agentlink/credential.agentlink/session.agentlink/status.json.claude/projects.codex/sessions

Network endpoints1

wss://link.harmopath.com

Decision evidence

public snapshot

AI called this Suspicious at 88.0% confidence as Dangerous Capability with low false-positive risk.

Evidence for warning

The user-invoked CLI starts a daemon that accepts run messages from a WebSocket hub.
A hub-supplied prompt is passed to detected coding-agent adapters, enabling remote agent execution in a selected working directory.
The daemon reads Claude and Codex configuration and session archives to support remote session access.
Remote execution approval is optional rather than enabled by default.

Evidence against

The manifest has no install, preinstall, or postinstall lifecycle hook.
The remote-control behavior is described in the package README and is activated by running the CLI.
Credentials and local state are written with owner-only permissions.
No hidden payload download, eval, or foreign AI-agent configuration write was found.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNativeBindingsNetworkShellWebSocket

Source & flagged code

10 flagged

dist/NodeSocket-CLfrrutd.mjsView file

Lines 4844-4884javascript

4844 parent: getOrUndefined(fromHeaders(request.headers)),

4845 kind: "server"

4846 });

4847 const prevServices = fiber.context;

4848 fiber.setContext(add(fiber.context, ParentSpan, span));

4849 return onExitPrimitive(httpApp, (exit) => {

4850 fiber.setContext(prevServices);

4851 const endTime = fiber.getRef(Clock).currentTimeNanosUnsafe();

4852 fiber.currentDispatcher.scheduleTask(() => {

4853 let response;

4854 let spanExit = exit;

4855 if (isFailure(exit)) {

4856 const [failureResponse, cause] = causeResponseStripped(exit.cause);

4857 response = failureResponse;

4858 spanExit = isSome(cause) ? failCause(cause.value) : succeed$4(response);

4859 } else response = exit.value;

4860 if (span.sampled) {

4861 const url = toURL(request);

4862 if (isSome(url) && (url.value.username !== "" || url.value.password !== "")) {

4863 url.value.username = "REDACTED";

4864 url.value.password = "REDACTED";

Medium

Secret Pattern

Package contains a possible secret pattern.

dist/NodeSocket-CLfrrutd.mjsView on unpkg · L4864

4865 }

4866 const redactedHeaderNames = fiber.getRef(CurrentRedactedNames);

4867 const requestHeaders = redact(request.headers, redactedHeaderNames);

4868 span.attribute("http.request.method", request.method);

4869 if (isSome(url)) {

4870 span.attribute("url.full", url.value.toString());

4871 span.attribute("url.path", url.value.pathname);

4872 if (url.value.search.slice(1) !== "") span.attribute("url.query", url.value.search.slice(1));

4873 span.attribute("url.scheme", url.value.protocol.slice(0, -1));

4874 }

4875 if (request.headers["user-agent"] !== void 0) span.attribute("user_agent.original", request.headers["user-agent"]);

4876 for (const name in requestHeaders) span.attribute(`http.request.header.${name}`, String(requestHeaders[name]));

4877 if (isSome(request.remoteAddress)) span.attribute("client.address", request.remoteAddress.value);

4878 span.attribute("http.response.status_code", response.status);

4879 const responseHeaders = redact(response.headers, redactedHeaderNames);

4880 for (const name in responseHeaders) span.attribute(`http.response.header.${name}`, String(responseHeaders[name]));

4881 }

4882 span.end(endTime, spanExit);

4883 }, 0);

4884 }, true);

bin/agentlink.jsView file

1#!/usr/bin/env node L2: import { spawn } from "node:child_process"; L3: import { randomBytes } from "node:crypto"; L4: import { createServer } from "node:http"; L5: import { existsSync } from "node:fs"; ... L28: } L29: const saved = JSON.parse(raw); L30: return saved.hub === normalizeHub(hub) && typeof saved.jwt === "string" ? saved.jwt : undefined; ... L59: L60: const OFFICIAL_HUB = process.env.AGENTLINK_DEFAULT_HUB || "wss://link.harmopath.com"; L61: const STATE_DIR = process.env.AGENTLINK_STATE_DIR || join(homedir(), ".agentlink"); L62: const CREDENTIAL_FILE = join(STATE_DIR, "credential");

Critical

Remote Asset Decode Execute

Source fetches a remote non-code asset, decodes its contents, and dynamically executes the decoded payload.

bin/agentlink.jsView on unpkg · L1

•Trigger-reachable chain: manifest.bin -> bin/agentlink.js Reachable file contains a blocking source-risk pattern.

Critical

Trigger Reachable Dangerous Capability

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

bin/agentlink.jsView on unpkg

1#!/usr/bin/env node L2: import { spawn } from "node:child_process"; L3: import { randomBytes } from "node:crypto";

High

Child Process

Package source references child process execution.

bin/agentlink.jsView on unpkg · L1

dist/ui.jsView file

•matchType = normalized_sha256 matchedPackage = alink-cli@0.10.0 matchedPath = dist/ui.js matchedIdentity = npm:YWxpbmstY2xp:0.10.0 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High

Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

dist/ui.jsView on unpkg

317"use strict"; L318: if (process.env.NODE_ENV !== "production") { L319: (function() { ... L545: isMounted: ["isMounted", "Instead, make sure to clean up subscriptions and pending requests in componentWillUnmount to prevent memory leaks."], L546: replaceState: ["replaceState", "Refactor your code to use setState instead (see https://github.com/facebook/react/issues/3236)."] L547: }; ... L2209: ]; L2210: if (process.platform !== "win32") { L2211: module.exports.push( ... L2357: } L2358: process13.exitCode = code || /* istanbul ignore next */ L2359: 0;

Medium

Install Persistence

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

dist/ui.jsView on unpkg · L317

dist/NodeServices-BtcL7AtO.mjsView file

•matchType = normalized_sha256 matchedPackage = alink-cli@0.10.0 matchedPath = dist/NodeServices-BtcL7AtO.mjs matchedIdentity = npm:YWxpbmstY2xp:0.10.0 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High

Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

dist/NodeServices-BtcL7AtO.mjsView on unpkg

dist/daemon.jsView file

•matchType = normalized_sha256 matchedPackage = alink-cli@0.10.0 matchedPath = dist/daemon.js matchedIdentity = npm:YWxpbmstY2xp:0.10.0 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High

Known Malware Source Similarity

Source file is highly similar to a previously finalized malicious package; route for source-aware review.

dist/daemon.jsView on unpkg

Source & flagged code

10 flagged · loading source

dist/NodeSocket-CLfrrutd.mjsView file

4864patternName = generic_password severity = medium line = 4864 matchedText = url.valu...ED";

Medium

Secret Pattern

Package contains a possible secret pattern.

dist/NodeSocket-CLfrrutd.mjsView on unpkg · L4864

bin/agentlink.jsView file

Critical

Remote Asset Decode Execute

Source fetches a remote non-code asset, decodes its contents, and dynamically executes the decoded payload.

bin/agentlink.jsView on unpkg · L1

Findings

2 Critical6 High6 Medium4 Low

CriticalRemote Asset Decode Executebin/agentlink.js

CriticalTrigger Reachable Dangerous Capabilitybin/agentlink.js

HighChild Processbin/agentlink.js

HighShell

HighCross File Remote Execution Contextbin/agentlink.js

HighKnown Malware Source Similaritydist/ui.js

HighKnown Malware Source Similaritydist/NodeServices-BtcL7AtO.mjs

HighKnown Malware Source Similaritydist/daemon.js

MediumSecret Patterndist/NodeSocket-CLfrrutd.mjs

MediumDynamic Requirebin/agentlink.js

MediumNetwork

MediumEnvironment Vars

MediumInstall Persistence

Affected versions and remediation

This report applies to alink-cli@0.11.14.

See all versions of alink-cli for other recorded verdicts.

Review the evidence and your use of alink-cli@0.11.14 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 04:43:10.435Z (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