registry / @raolin2025/claude-code-node / 3.6.15

@raolin2025/claude-code-node@3.6.15

Node.js AI Code Agent CLI - Zero dependencies, pure JavaScript, security hardened, multi-channel notifications, Telegram & QQ Bot remote programming, rich media upload, multi-account management

AI Security Review

scanned 12d ago · by lpm-firewall-ai

The inspected source ties the flagged requests to configured Telegram functionality and an explicit npm publishing operation targeting the npm registry. The manifest has no install lifecycle hook, and the reviewed evidence does not establish an active attack.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

Running the notifier or explicitly invoking its publishing tool activates the relevant operations.

Impact

No package-controlled credential exfiltration or unconsented install-time behavior was established.

Mechanism

Telegram requests use channel configuration; npm publishing sends the token to the npm registry endpoint.

AI rationale

Decision evidence

public snapshot

AI called this Clean at 84.0% confidence as Benign with low false-positive risk.

Evidence for AI clean decision

The manifest exposes CLI entrypoints and a prepublish-only check, with no install lifecycle hook.
The notifier sends Telegram typing requests to the configured Telegram API using the configured channel token.
The publish tool reads an npm token and uses it for publishing to the npm registry as an explicit tool action.
The inspected security code applies permission checks and blocks DNS results resolving to private addresses.

Evidence against

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShellWebSocket

Supply chain

HighEntropyStringsPossibleObfuscationUrlStrings

ManifestNo manifest risk signals triggered.

scanned 76 file(s), 779 KB of source, external domains: 127.0.0.1, api.deepseek.com, api.github.com, api.search.brave.com, api.sgroup.qq.com, api.telegram.org, bots.qq.com, r.jina.ai, registry.npmjs.org, www.googleapis.com

Source & flagged code

11 flagged

src/tools/bash.jsView file

Lines 1-31javascript

1/**

2 * Bash 工具 — 执行 shell 命令（跨平台）

3 * 对应原版: src/tools/BashTool/

4 *

5 * 跨平台：通过 src/utils/shell.js 探测层自动选择可用 shell。

6 * - Linux/macOS → /bin/bash（或 SHELL）

7 * - Windows (Git Bash) → Git Bash（bash 语义）

8 * - Windows (WSL) → wsl bash（bash 语义）

9 * - Windows (PS) → PowerShell

10 * - Windows (cmd) → cmd.exe 兜底

11 */

12import { spawn } from 'child_process'

13import { ToolDef } from '../types/index.js'

High

Child Process

Package source references child process execution.

src/tools/bash.jsView on unpkg · L11

14import { checkBashSafety } from '../security/bash-guard.js'

15import { detectShell, getShellDescription } from '../utils/shell.js'

17// 启动时探测一次并缓存

18const shellCfg = detectShell()

20/**

21 * 动态生成工具描述，告知 LLM 当前平台的 shell 语义，

22 * 避免模型在 Windows 上写出 Linux 命令导致失败。

23 */

24function buildDescription() {

25 const envHint = getShellDescription()

26 return `Execute a shell command. The command will run in a shell subprocess.

28${envHint}

30Usage:

31- Provide the command as the value of the 'command' key (use the syntax matching the current shell above)

src/core/scheduler.jsView file

Lines 306-346javascript

306 const raw = readFileSync(execLogFile, 'utf-8')

307 const lines = raw.split('\n')

308

src/tools/loadCustomTools.jsView file

Lines 115-155javascript

115 )

116 }

117

118 return null

119}

120

121/** 从模块导出里收集所有候选原始工具定义 */

122function collectExports(mod) {

123 const

src/tools/npm-publish.jsView file

Lines 1-24javascript

1/**

2 * NpmPublish - npm 发布一键工具

3 *

4 * 封装了本项目 npm 发布的全部经验（见 NPM_STAGED_PUBLISH_GUIDE.md）：

5 * - 标准 `npm publish` 常因 npm CLI 强制 otplease/2FA 返回 403

6 * - 最有效方法：bypass-2FA GAT token + 手动构造 PUT 请求 + `npm-auth-type: bearer`，无需 OTP

7 *

8 * 功能：

9 * - version 升版本号（patch/minor/major 或指定版本）

10 * - status 检查登录 / 当前版本 / registry 版本 / 暂存区状态

src/channel/notify-daemon.jsView file

Lines 2-50javascript

2 * cc-notify — 通知守护进程（v2.0 增强版）

3 *

4 * 架构：

5 * Telegram ← Bot API → cc-notify (长轮询) → cc-node

src/security/ssrf-guard.jsView file

Lines 1-28javascript

1/**

2 * SSRF 防护 — 阻止对内网/云元数据端点的请求

3 * 对应原版：src/utils/hooks/ssrfGuard.ts

4 *

5 * v1.2 修复:

6 * - H4: 完整检测 fe80::/10 范围（fe80:: - febf::），而非仅 fe8 前缀

7 * - H3: 导出 dnsLookup 函数供 query-engine 使用自定义 DNS resolver

8 */

9import { lookup as dnsLookup } from 'dns'

src/security/enhanced-permission.jsView file

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. ult.reasons.join('; ') : undefined, detail: result, } } case 'webfetch': { const url = input.url || '' if (!url) return { safe: true } const result = await checkurlsafety(url) return { safe: result.allowed, reason: result.reason, } } default: return { safe: true } } } _extractpattern(toolname, input) { switch (toolname) { case 'bash': return input.command || '' case 'read': case 'edit': case 'write': return input.file_path || '' case 'glob': case 'grep':

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.

src/security/enhanced-permission.jsView on unpkg

Source & flagged code

11 flagged · loading source

src/tools/bash.jsView file

11*/ L12: import { spawn } from 'child_process' L13: import { ToolDef } from '../types/index.js'

High

Child Process

Package source references child process execution.

src/tools/bash.jsView on unpkg · L11

src/core/scheduler.jsView file

326try { L327: child = spawn(cmd, { shell: true, cwd, timeout, env: { ...process.env, ...(task.action?.env || {}) } }) L328: } catch (e) { result.error = String(e.message || e); return resolve() }

High

Shell

Package source references shell execution.

src/core/scheduler.jsView on unpkg · L326

src/tools/loadCustomTools.jsView file

135async function loadToolFile(filePath, fallbackName, log) { L136: const mod = await import(pathToFileURL(filePath).href) L137: const raws = collectExports(mod)

Findings

4 Critical5 High4 Medium6 Low

CriticalSame File Env Network Executionsrc/channel/notify-daemon.js

CriticalHardcoded Runtime Data Exfiltrationsrc/channel/notify-daemon.js

CriticalCredential Exfiltrationsrc/tools/npm-publish.js

CriticalTrigger Reachable Dangerous Capabilitysrc/channel/notify-daemon.js

HighChild Processsrc/tools/bash.js

HighShellsrc/core/scheduler.js

HighCloud Metadata Accesssrc/security/ssrf-guard.js

HighTrigger Reachable External Ai Agent Control Surface Mutationsrc/security/enhanced-permission.js

HighTrigger Reachable Credential Exfiltrationsrc/channel/notify-daemon.js

MediumDynamic Requiresrc/tools/loadCustomTools.js

Medium

Affected versions and remediation

This report applies to @raolin2025/claude-code-node@3.6.15.

See all versions of @raolin2025/claude-code-node for other recorded verdicts.

Review the evidence and your use of @raolin2025/claude-code-node@3.6.15 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: 2026-10-11 07:20:34.250Z (UTC).

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