registry / @tiens.nguyen/gu-cli / 1.0.754

@tiens.nguyen/gu-cli@1.0.754

Gu CLI — the gu terminal plus the local worker that runs agent / OCR / PDF / embedding jobs on your Mac (Ollama / MLX / OpenAI-compatible).

AI Security Review

scanned 14d ago · by lpm-firewall-ai

Source inspection found a narrow, package-aligned PATH update and an interactive worker-key flow that validates against the selected API before saving. The flagged MCP location is part of project configuration handling, and the inspected source describes approval before project servers start; no malicious attack surface was established.

Static reason

One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source

Trigger

Package installation triggers the PATH setup; interactive REPL use triggers sign-in or worker-key entry.

Impact

The inspected behavior enables command discovery and worker authentication; no concrete malicious effect was found.

Mechanism

The install hook may append a marked PATH block when the global bin directory is missing. A pasted worker key is checked against the configured API before it is saved.

AI rationale

Evidence

install-path.mjs~/.gonext/worker.env

Network endpoints1

chat-api-v2.gomarsic.cc

Decision evidence

public snapshot

AI called this Clean at 87.0% confidence as Benign with low false-positive risk.

Evidence for AI clean decision

package.json’s postinstall runs install-path.mjs; that script adds the global bin directory to shell PATH only when needed.
The REPL asks the user to paste an existing worker key or choose browser sign-in; saving a pasted key validates it with a read-only request to the selected API base before storing it.
The default API base is the package’s Gu service; the source shows no forwarding of this worker key to an unrelated destination.
The MCP configuration exposes a project configuration path; project servers are gated on approval before they start.
Client Expert starts its local API using the current Node executable and a specified API directory.
The remaining required source paths describe provider setup, proxy routing, and graceful child-process stopping; no malicious behavior was identified in the inspected excerpts.

Evidence against

Static hints flag credential forwarding, shell execution, and install-time behavior, but the inspected code shows a user-entered worker key validated against the chosen API and a narrow PATH update.
The package defines a postinstall hook, but the inspected target does not mutate an AI-agent control surface.

Source & flagged code

12 flagged

package.jsonView file

Lines 1-41json

2 "name": "@tiens.nguyen/gu-cli",

3 "version": "1.0.754",

4 "description": "Gu CLI — the gu terminal plus the local worker that runs agent / OCR / PDF / embedding jobs on your Mac (Ollama / MLX / OpenAI-compatible).",

5 "type": "module",

6 "license": "MIT",

7 "repository": {

8 "type": "git",

9 "url": "git+https://github.com/tiennsloit/gu.git"

10 },

11 "homepage": "https://github.com/tiennsloit/gu#readme",

12 "bugs": {

13 "url": "https://github.com/tiennsloit/gu/issues"

14 },

15 "bin": {

16 "gu-cli": "gu-cli.mjs",

17 "gu-local-worker": "gu-cli.mjs",

18 "gu": "gu-repl.mjs"

19 },

20 "scripts": {

21 "postinstall": "node install-path.mjs",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L21

22 "test": "python3 -m unittest discover tests -v && node --test",

23 "test:js": "node --test",

24 "test:keys": "node --test tests/key-choice-regression.test.mjs tests/coding-key-line.test.mjs tests/coding-key-picker.test.mjs tests/coding-key-prompt.test.mjs tests/coding-key-notice.test.mjs tests/coding-key-fallback-line.test.mjs tests/manage-coding-providers.test.mjs tests/add-coding-provider.test.mjs tests/coding-provider-setup.test.mjs",

25 "deploy:local": "npm install -g .",

26 "run": "gu-cli",

27 "publish:org": "npm publish --access public",

28 "preflight": "node scripts/preflight.mjs",

29 "test:e2e": "GONEXT_E2E=1 node --test tests/client-expert-e2e.test.mjs",

30 "test:linux": "bash scripts/e2e-linux.sh",

31 "test:rag": "node scripts/rag-e2e.mjs",

32 "test:ollama1-e2e": "GONEXT_OLLAMA1_LIVE=1 node --test --test-timeout=1800000 tests/ollama1-e2e.test.mjs",

33 "test:ollama1": "GONEXT_OLLAMA1_LIVE=1 python3 -m unittest discover tests -p \"test_ollama1_live.py\" -v",

34 "test:proxy": "python3 -m unittest discover tests -p \"test_proxy.py\" && node --test --test-timeout=30000 tests/proxy-config.test.mjs tests/proxy-transport.test.mjs tests/proxy-local-bypass.test.mjs tests/proxy-selftest.test.mjs tests/proxy-command.test.mjs",

35 "coverage": "node scripts/coverage.mjs",

36 "capture:cache": "GONEXT_OLLAMA1_LIVE=1 node scripts/capture-cache-fixture.mjs",

37 "test:cache-e2e": "GONEXT_OLLAMA1_LIVE=1 node --test --test-timeout=1800000 tests/prefix-cache-e2e.test.mjs",

38 "test:tokens": "python3 -m unittest discover tests -p \"test_cached_tokens.py\" -v && node --test tests/prefix-cache.test.mjs tests/agent-turn-shape.test.mjs",

39 "test:selection": "node --test tests/selection-scrollback.test.mjs tests/selection-while-thinking.test.mjs tests/terminal-selection.test.mjs tests/terminal-copy.test.mjs tests/terminal-viewport.test.mjs tests/hover-repaint.test.mjs",

40 "test:steps": "python3 scripts/py-tap.py tests/test_read_caps.py tests/test_trim_payback.py tests/test_explore.py tests/test_and_chains.py tests/test_edit_file_primary.py tests/test_bare_code_rescue.py tests/test_batching_rule.py tests/test_no_pdf_routing.py",

41 "check:image": "node scripts/image-paste-check.mjs",

proxy-selftest.mjsView file

Lines 175-215javascript

175 proxyFx.seen.length = 0;

176 const off = { ...proxy, enabled: false };

177 if (shouldProxy(`${MODEL_URL}/chat/completions`

gu-cli.mjsView file

Lines 1-38javascript

1#!/usr/bin/env node

2/**

3 * Gu local worker:

4 * - `gu-cli set <workerKey> [--api-base URL]`

5 * writes ~/.gonext/worker.env

6 * - `gu-cli ws-ping-test` — POST worker ws-ping (needs GONEXT_* env)

7 * - `gu-cli simulate-chat [text]` — claim next chat job, push fake reply like the real worker (needs GONEXT_* env)

8 * - `gu-cli embed --model <mlx-embed-model> [--port 8085]` — run the MLX embeddings server (RAG); use a separate terminal

9 * - `gu-cli workspace add|remove|list …` — manage local code folders the agent may read/edit/test in

10 * - `gu-cli` — starts polling loop (claims jobs and runs models)

client-expert-api.mjsView file

Lines 154-194javascript

154 /**

155 * The shell form: ONE string, no args array.

156 *

157 * Node 22 prints DEP0190 — "Passing args to a child process with shell option true can

158 * lead to security vulnerabilities, as the arguments are not escaped, only concatenated" —

159 * and printed it in the middle of the Windows setup output, which reads like something

160 * went wrong. Node is right: with a shell, the array is concatenated anyway, so building

161 * the line explicitly is both quieter AND more honest about what runs. Safe because every

162 * argument here is a literal flag (asserted in the tests).

rag-selftest.mjs#virtual:normalized:round1View file

34function haystackFile(n) { L35: const lines = [`// Service module ${n}: request handling for the platform"s internal API.`];\n for (let i = 1; i <= 40; i++) {\n lines.push(\n `// helper ${i}: validates t... L36: // embedder is down" when it is simply a different path.\n const raw = String(s.ragEmbedUrl || "").replace(//+$/, "");\n embedUrl = raw && !//v1$/i.test(raw) ? `${raw}/v1` : ...

Medium

Dynamic Require

Package source references dynamic require/import behavior.

rag-selftest.mjs#virtual:normalized:round1View on unpkg · L34

gu-repl.mjsView file

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. gu-repl.mjs: import { readFile, mkdir, writeFile, unlink, rename } from "node:fs/promises"; import { existsSync, statSync, readFileSync, writeFileSync, readdirSync, openSync, readSync, closeSync } from "node:fs"; import { createInterface } from "node:readline"; // A SELF-CONTAINED prompt: this runs before the REPL's readline (and its ask()) is // ONE readline for the whole first run: the mode question and the key prompt share it, // the REPL's own readline exists, so it owns the terminal here). const rl0 = (await import("node:readline")).createInterface({ // Exit AFTER the readline is closed, so the terminal is left in a usable state.

High

Credential Redirect Persistence

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

gu-repl.mjsView on unpkg

coding-provider-setup.mjsView file

Lines 1-32javascript

1/**

2 * Adding an OpenAI-compatible coding provider from the terminal (task #223).

3 *

4 * node --test tests/coding-provider-setup.test.mjs

5 *

6 * The DECISIONS live here and the presentation lives in gu-repl.mjs, the same split

7 * ollama-setup.mjs / reconfigureOllamaServer uses — so what a URL means and what a model list is

8 * worth can be tested without a terminal, and first-run setup and /model cannot disagree.

9 *

10 * WHY THE URL RULES ARE NOT ollama-setup's. normalizeOllamaUrl DISCARDS the path, and it is right

proxy-dispatcher.mjsView file

Lines 21-61javascript

21 * Not an optimisation — a new agent per request opens a new connection pool per request, and a

22 * proxy that authenticates then re-authenticates on every single call, which is both slow and a

23 * good way to trip an account lockout policy. Keyed by the dial string so changing the password

24 * produces a new agent rather than silently reusing the old credentials.

gonext_transcribe.pyView file

•path = gonext_transcribe.py kind = build_helper sizeBytes = 4248 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

gonext_transcribe.pyView on unpkg

agent-stop.mjs#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 5

High

Semantic Analysis Limited

A bounded semantic-analysis stage reached its safety limit; remaining detectors completed, but this package requires AI review.

agent-stop.mjs#virtual:normalized:round1View on unpkg

Source & flagged code

12 flagged · loading source

package.jsonView file

•scripts.postinstall = node install-path.mjs

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

proxy-selftest.mjsView file

195patternName = generic_password severity = medium line = 195

Medium

Secret Pattern

Package contains a possible secret pattern.

proxy-selftest.mjsView on unpkg · L195

gu-cli.mjsView file

18import { mkdir, mkdtemp, readFile, rm, stat, unlink, writeFile } from "node:fs/promises"; L19: import { execFile as execFileCallback, spawn } from "node:child_process"; L20: import { createHash } from "node:crypto";

High

Findings

9 High6 Medium4 Low

HighInstall Time Lifecycle Scriptspackage.json

HighChild Processgu-cli.mjs

HighShellclient-expert-api.mjs

HighSame File Env Network Executiongu-cli.mjs

HighCredential Redirect Persistencegu-repl.mjs

HighCloud Metadata Accesscoding-provider-setup.mjs

HighCopied Package Dependency Bridgeproxy-dispatcher.mjs

HighSemantic Analysis Limitedagent-stop.mjs#virtual:normalized:round1

HighPrevious Version Dangerous Deltaclient-expert-api.mjs

MediumSecret Patternproxy-selftest.mjs

MediumDynamic Requirerag-selftest.mjs#virtual:normalized:round1

MediumNetwork

Affected versions and remediation

This report applies to @tiens.nguyen/gu-cli@1.0.754.

See all versions of @tiens.nguyen/gu-cli for other recorded verdicts.

Review the evidence and your use of @tiens.nguyen/gu-cli@1.0.754 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: 2026-10-09 05:21:26.954Z (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

Behavioral surface

Source & flagged code