Loading npm security reports…

@babylonjs-toolkit/agent@1.1.16: AI-agent control-surface risk | LPM Firewall

registry / @babylonjs-toolkit/agent / 1.1.16

@babylonjs-toolkit/agent@1.1.16

Installs the Babylon Toolkit agent skills and Agent Persona for Claude Code, Codex, GitHub Copilot, Gemini CLI and Antigravity.

AI Security Review

scanned 27d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. On global npm installation, the postinstall hook writes skills and instructions into several AI-agent control surfaces. It also enables Codex workspace network access and installs instructions that require a remote document to be trusted.

Static reason

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

Trigger

Running npm global installation with lifecycle scripts enabled.

Impact

Future agent sessions can inherit package-controlled instructions and remote content with expanded network capability.

Mechanism

Automatic mutation of global AI-agent instructions, skills, and Codex network settings.

Policy narrative

The package runs a postinstall hook during global npm installation. That hook invokes the installer, which writes into global instruction and skill locations for multiple unrelated AI clients. For Codex it creates or changes configuration to turn network access on. The installed persona then directs future agents to fetch and treat a remote GitHub document as authoritative before handling Babylon-related work. This is an unconsented lifecycle mutation of broad agent control surfaces.

AI rationale

The global postinstall path automatically modifies multiple AI-agent instruction surfaces and enables Codex network access. The remote-authority instruction makes the resulting control path persistent across later agent sessions.

Evidence

package.jsonscripts/postinstall.jslib/targets.jslib/codex-config.jspersona.md~/.claude/skills

Source & flagged code

10 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

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 1

High

Semantic Analysis Limited

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

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

lib/selfupdate.jsView file

4const path = require('path'); L5: const { execFileSync } = require('child_process'); L6:

High

Child Process

Package source references child process execution.

lib/selfupdate.jsView on unpkg · L4

207const script = ` L208: const https = require('https'); L209: const url = 'https://registry.npmjs.org/' + encodeURIComponent(${JSON.stringify(PACKAGE_NAME)}).replace('%40','@') + '/latest'; ... L213: res.on('end', () => { L214: try { process.stdout.write(String(JSON.parse(body).version || '')); } catch { process.stdout.write(''); } L215: }); ... L220: try { L221: const out = execFileSync(process.execPath, ['-e', script], { encoding: 'utf8', timeout: timeout + 2000 }); L222: return String(out).trim() || null;

High

Command Output Exfiltration

Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.

lib/selfupdate.jsView on unpkg · L207

207Trigger-reachable command-output exfiltration chain: manifest.bin -> bin/bt-agent.js -> lib/selfupdate.js L207: const script = ` L208: const https = require('https'); L209: const url = 'https://registry.npmjs.org/' + encodeURIComponent(${JSON.stringify(PACKAGE_NAME)}).replace('%40','@') + '/latest'; ... L213: res.on('end', () => { L214: try { process.stdout.write(String(JSON.parse(body).version || '')); } catch { process.stdout.write(''); } L215: }); ... L220: try { L221: const out = execFileSync(process.execPath, ['-e', script], { encoding: 'utf8', timeout: timeout + 2000 }); L222: return String(out).trim() || null;

High

Trigger Reachable Command Output Exfiltration

A manifest entrypoint or package-local install chain reaches command-output exfiltration behavior.

lib/selfupdate.jsView on unpkg · L207

•matchType = normalized_sha256 matchedPackage = @babylonjs-toolkit/agent@1.1.11 matchedPath = lib/selfupdate.js matchedIdentity = npm:[redacted]:1.1.11 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.

lib/selfupdate.jsView on unpkg

•matchType = malicious_source_fingerprint_signature signature = 830d844af9032a30 signatureType = suspicious_hashes sourceLabel = final_verdict:malicious matchedPackage = @babylonjs-toolkit/agent@1.1.11 matchedPath = lib/selfupdate.js matchedIdentity = npm:[redacted]:1.1.11 similarity = 1.000 shingleOverlap = 13 summary = package final verdict is malicious

High

Known Malware Source Fingerprint Signature

Source fingerprint signature matches a known malicious package signature; route for source-aware review.

lib/selfupdate.jsView on unpkg

lib/targets.jsView file

1Install-time code directly mutates a foreign AI-agent control surface: L20: default: true, L21: global: { skills: '~/.claude/skills', instructions: '~/.claude/CLAUDE.md' }, L22: project: { skills: '.claude/skills', instructions: 'CLAUDE.md' }, L23: }, ... L27: default: true, L28: global: { skills: '~/.agents/skills', instructions: '~/.agents/AGENTS.md' }, L29: project: { skills: '.agents/skills', instructions: 'AGENTS.md' }, L30: }, ... L34: default: true, L35: global: { skills: null, instructions: '~/.codex/AGENTS.md', config: '~/.codex/config.toml' }, L36: project: { skills: null, instructions: 'AGENTS.md', config: '.codex/config.toml' }, L37: }, Write operation from lib/codex-config.js: L80: } L81: const template = fs.readFileSync(path.join(PACKAGE_ROOT, '.codex', 'config.toml'), 'utf8'

Critical

Ai Agent Control Hijack

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

lib/targets.jsView on unpkg · L1

skills/bt-atlas/scripts/preview.pyView file

•path = skills/bt-atlas/scripts/preview.py kind = build_helper sizeBytes = 2478 magicHex = [redacted]

Medium

Ships Build Helper

Package ships non-JavaScript build or shell helper files.

skills/bt-atlas/scripts/preview.pyView on unpkg

Lines 187-227javascript

187function fetchLatestVersion({ timeout = 20000 } = {}) {

188 try {

189 const out = runManager('npm', ['view', PACKAGE_NAME, 'version'], {

190 encoding: 'utf8',

191 timeout,

192 stdio: ['ignore', 'pipe', 'pipe'],

193 });

194 const v = String(out).trim().split(/\s+/).pop();

195 if (v) return { version: v };

196 } catch (err) {

197 const reason = (err && (err.stderr || err.message) ? String(err.stderr || err.message) : '').trim();

198 const fallback = fetchLatestFromRegistrySync(timeout);

199 if (fallback) return { version: fallback };

200 return { error: reason.split('\n').filter(Boolean).pop() || 'could not reach the npm registry' };

201 }

202 return { error: 'npm returned no version' };

203}

204

205/** Registry lookup without the npm CLI. Runs in a helper process so it can be sync. */

206function fetchLatestFromRegistrySync(timeout) {

207 const script = `

208 const https = require('https');

209 const url = 'https://registry.npmjs.org/' + encodeURIComponent(${JSON.stringify(PACKAGE_NAME)}).replace('%40','@') + '/latest';

210 const req = https.get(url, { headers: { accept: 'application/json' } }, (res) => {

211 let body = '';

212 res.on('data', (c) => { body += c; });

213 res.on('end', () => {

214 try { process.stdout.write(String(JSON.parse(body).version || '')); } catch { process.stdout.write(''); }

215 });

216 });

217 req.setTimeout(${Math.max(1000, timeout)}, () => req.destroy());

218 req.on('error', () => process.stdout.write(''));

219 `;

220 try {

221 const out = execFileSync(process.execPath, ['-e', script], { encoding: 'utf8', timeout: timeout + 2000 });

222 return String(out).trim() || null;

High

Command Output Exfiltration

Source combines command execution, command-output handling, and outbound requests; review data flow before blocking.

lib/selfupdate.jsView on unpkg · L207

High

Trigger Reachable Command Output Exfiltration

A manifest entrypoint or package-local install chain reaches command-output exfiltration behavior.

lib/selfupdate.jsView on unpkg · L207

223 } catch {

224 return null;

225 }

226}

227