Loading npm security reports…

@su-record/vibe@4.1.4: AI-agent control-surface risk | LPM Firewall

registry / @su-record/vibe / 4.1.4

@su-record/vibe@4.1.4

An AX/FDE harness for Claude Code, Codex CLI and ChatGPT desktop. Say what you need, approve the scenarios once; the harness proves the work by running the checks itself. Memory in plain files, human tokens for irreversible actions, a ledger instead of cl

AI Security Review

scanned 3h ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. A global npm postinstall alters Claude, Codex, and Hermes agent home directories. It injects instructions, replaces skill directories, and registers commands on Bash and edit events.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

Global installation of @su-record/vibe runs its postinstall script; later CLI commands repair the changes.

Impact

The package can persistently influence agent behavior and cause package code to run on future agent tool events without an explicit setup action.

Mechanism

Install-time mutation of broad AI-agent instruction and hook configurations.

Policy narrative

On global installation, npm automatically invokes the package postinstall script. That script calls global setup, which writes agent instruction files, removes and replaces skill directories, and installs commands into Claude and Codex hook configurations. Those hooks are activated for Bash and edit-related agent events. Subsequent ordinary CLI commands silently restore missing surfaces, making the change persistent. This is unconsented mutation of broad, foreign AI-agent control surfaces.

Rationale

The lifecycle-triggered global setup directly modifies multiple AI-agent home configurations and installs event-driven commands. This meets the install-hook abuse policy regardless of the unrelated, user-invoked GitHub functionality.

Evidence

scripts/postinstall.jsdist/install/global.jshooks/notify.jsdist/cli.js~/.claude/CLAUDE.md~/.claude/settings.json~/.claude/skills/vibe~/.codex/AGENTS.md

Source & flagged code

9 flagged · loading source

package.jsonView file

•scripts.postinstall = node scripts/postinstall.js

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

•scripts.postinstall = node scripts/postinstall.js

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg

mcpb/server/index.jsView file

10*/ L11: import { spawnSync } from 'node:child_process'; L12: import fs from 'node:fs';

High

Child Process

Package source references child process execution.

mcpb/server/index.jsView on unpkg · L10

dist/core/checks/run.jsView file

18return new Promise((resolve) => { L19: const child = spawn(check.cmd, { cwd, shell: true, env: { ...process.env, VIBE_CHECK: '1' }, stdio: ['ignore', 'pipe', 'pipe'] }); L20: let captured = '';

High

Shell

Package source references shell execution.

dist/core/checks/run.jsView on unpkg · L18

dist/core/tree.jsView file

1import { createHash } from 'node:crypto'; L2: import { spawnSync } from 'node:child_process'; L3: import fs from 'node:fs'; ... L7: const result = spawnSync('git', args, { cwd: root, encoding: 'utf-8' }); L8: return result.status === 0 ? result.stdout : null; L9: }

Low

Weak Crypto

Package source references weak cryptographic algorithms.

dist/core/tree.jsView on unpkg · L1

dist/core/github.jsView file

1import { spawnSync } from 'node:child_process'; L2: import { readJson } from './store.js'; L3: import { usage } from './errors.js'; L4: const API = 'https://api.github.com'; L5: export function githubToken() { L6: const env = process.env['GITHUB_TOKEN'] ?? process.env['GH_TOKEN']; L7: if (env)

High

Same File Env Network Execution

A single source file combines environment access, network access, and code or shell execution; review context before blocking.

dist/core/github.jsView on unpkg · L1

dist/install/global.jsView file

11export const SKILL_NAMES = ['vibe', 'vibe.discover', 'vibe.scope', 'vibe.build', 'vibe.prove', 'vibe.handoff']; L12: const CLIENT_DIR = { claude: '.claude', codex: '.codex', hermes: '.hermes' }; L13: /** User-level layout: one per client home. Codex reads a native `~/.codex/hooks.json` shaped like Claude's settings. */ L14: export function globalLayout(client) { ... L16: if (client === 'claude') L17: return { card: path.join(dir, 'CLAUDE.md'), skills: path.join(dir, 'skills'), hook: path.join(dir, 'settings.json') }; L18: if (client === 'codex') L19: return { card: path.join(dir, 'AGENTS.md'), skills: path.join(dir, 'skills'), hook: path.join(dir, 'hooks.json') }; L20: // Hermes Agent: identity file SOUL.md is the only home-level instruction it always loads; skills follow the agentski

Critical

Ai Agent Control Hijack

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

dist/install/global.jsView on unpkg · L11

•Manifest-trigger-reachable source links an external AI-agent control path to a behavior-bearing write operation. import fs from 'node:fs'; import os from 'node:os'; import path from 'node:path'; import { packageroot } from '../core/paths.js'; import { readjson, readtext, writeatomic, writejson } from '../core/store.js'; import { claudepluginversion, cliavailable, codexregistered, registerclaude, registercodex, unregisterclaude, unregistercodex } from './register.js'; export const all_clients = ['claude', 'codex', 'hermes']; export const card_start = ''; export const card_end = ''; export const card_max_bytes = 1024; export const skill_names = ['vibe', 'vibe.discover', 'vibe.scope', 'vibe.build', 'vibe.prove', 'vibe.handoff']; const client_dir = { claude:

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.

dist/core/docs/pdf.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 2

High

Semantic Analysis Limited

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

dist/core/docs/pdf.js#virtual:normalized:round1View on unpkg

Source & flagged code

9 flagged

package.jsonView file

Lines 9-49json

9 "files": [

10 "dist/",

11 "skills/",

12 "hooks/",

13 "scripts/",

14 ".claude-plugin/",

15 ".codex-plugin/",

16 "card.md",

17 "README.md",

18 "LICENSE",

19 "mcpb/"

20 ],

21 "engines": {

22 "node": ">=20.12.0"

23 },

24 "scripts": {

25 "build": "tsc -p tsconfig.json && chmod +x dist/cli.js",

26 "typecheck": "tsc --noEmit -p tsconfig.json",

27 "test": "vitest run",

28 "check": "npm run build && npm test",

29 "postinstall": "node scripts/postinstall.js"

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L29

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg · L29

30 },

31 "dependencies": {

32 "ajv": "^8.17.1",

33 "yaml": "^2.8.1"

34 },

35 "devDependencies": {

36 "@types/node": "^22.19.10",

37 "tsx": "^4.23.5",

38 "typescript": "^5.9.3",

39 "vitest": "^4.1.10"

40 },

41 "license": "MIT",

42 "repository": "https://github.com/su-record/vibe",

43 "keywords": [

44 "harness",

45 "ax",

46 "fde",

47 "forward-deployed-engineer",

48 "verification",

49 "claude-code",

mcpb/server/index.jsView file

Lines 1-30javascript

1#!/usr/bin/env node

2/**

3 * vibe for the Claude desktop app — an MCP Bundle server that is nothing but a cable to the

4 * `vibe` CLI. Every tool runs `vibe <command> --json` in the project folder the user picked at

5 * install time (VIBE_PROJECT_DIR) and returns the JSON. The verdict stays in the CLI; this

6 * surface is for the interview, the approval, the judgement and the handoff — not for writing code.

7 *

8 * MCP over stdio is newline-delimited JSON-RPC 2.0. Three methods matter here: initialize,

9 * tools/list, tools/call. No SDK: the protocol subset is small and the bundle stays dependency-free.

10

dist/core/checks/run.jsView file

Lines 1-38javascript

1import { spawn } from 'node:child_process';

2import path from 'node:path';

3const DEFAULT_TIMEOUT_MS = 600_000;

4const TAIL_LINES =

dist/core/tree.jsView file

Lines 1-21javascript

1import { createHash } from 'node:crypto';

2import { spawnSync } from 'node:child_process';

3import fs from 'node:fs';

4import path from 'node:path';

dist/core/github.jsView file

Lines 1-21javascript

1import { spawnSync } from 'node:child_process';

2import { readJson } from './store.js';

3import { usage } from './errors.js';

4const API =

dist/install/global.jsView file

Lines 1-31javascript

1import fs from 'node:fs';

2import os from 'node:os';

3import path from 'node:path';

4import { packageRoot } from '../core/paths.js';

dist/core/docs/pdf.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 2

High

Semantic Analysis Limited

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

dist/core/docs/pdf.js#virtual:normalized:round1View on unpkg