Loading npm security reports…

@su-record/vibe@4.1.5: AI-agent control-surface risk | LPM Firewall

registry / @su-record/vibe / 4.1.5

@su-record/vibe@4.1.5

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 automatically modifies AI-agent home directories. It adds instruction cards, skills, and commands triggered for Bash and tool-use events, then restores missing surfaces during ordinary CLI use.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

Installing the package globally; later running most vibe CLI commands can restore the changes.

Impact

The package can change the behavior and hook configuration of installed Claude, Codex, and Hermes clients without an explicit setup command.

Mechanism

Automatic AI-agent control-surface injection and self-repair.

Policy narrative

The package declares postinstall and, during a global npm installation, calls setupGlobal automatically. That routine targets installed AI-agent home directories, adding cards, skills, and hook configuration. The hooks launch the package notification command for Bash and editing events. The CLI also calls ensureGlobal for nearly every command, recreating stale or removed global surfaces. This is an unconsented lifecycle-driven modification of broad, foreign AI-agent control surfaces.

Rationale

The automatic postinstall path and subsequent self-repair create a concrete persistent AI-agent control-surface mutation. The global-only guard does not provide explicit user consent for those changes.

Evidence

package.jsonscripts/postinstall.jsdist/install/global.jsdist/cli.js~/.claude/CLAUDE.md~/.claude/skills~/.claude/settings.json~/.codex/AGENTS.md~/.codex/skills

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