Loading npm security reports…

@impakers/debug@1.9.8: AI-agent control-surface risk | LPM Firewall

registry / @impakers/debug / 1.9.8

@impakers/debug@1.9.8

Client feedback collection tool for Impakers OS

AI Security Review

scanned 10d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. An automatic install hook changes the consumer project's Claude agent control surface. No network attack was identified.

Static reason

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

Trigger

Installing the package through npm with lifecycle scripts enabled.

Impact

A package installation can cause Claude to automatically apply package-authored instructions during later development work.

Mechanism

The postinstall script copies a package-controlled skill into .claude/skills and also modifies an ESLint configuration when present.

Policy narrative

Installing the package automatically launches its setup script. That script writes a package-controlled Claude skill into the consuming project and invokes the operation during postinstall, rather than only after an explicit setup command. The injected skill is configured to activate for broad, ordinary UI development tasks. This is an unconsented lifecycle mutation of a foreign AI-agent control surface.

AI rationale

The package performs an automatic postinstall write to a consumer project's .claude skill directory. This concrete AI-agent control-surface mutation meets the publish-block policy.

Evidence

package.jsonscripts/setup.mjstemplates/ui-mapping-skill.md.claude/skills/impakers-ui-mapping/SKILL.mdeslint.config.mjseslint.config.jseslint.config.cjs

Decision evidence

Source & flagged code

5 flagged · loading source

package.jsonView file

•scripts.postinstall = node ./scripts/setup.mjs

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

scripts/setup.mjsView file

7// L8: // 1. .claude/skills/impakers-ui-mapping/SKILL.md 설치 L9: // 에이전트가 UI 코드를 짤 때 파일 매핑 규칙을 알고 있게 만든다. ... L114: /** L115: * .claude/skills/impakers-ui-mapping/SKILL.md 를 깐다. L116: * ... L126: L127: const target = path.join(root, ".claude", "skills", "impakers-ui-mapping", "SKILL.md"); L128: const next = fs.readFileSync(source, "utf-8"); ... L136: if (current === next) return; L137: fs.writeFileSync(target, next); L138: note(`스킬 갱신: ${rel(root, target)}`);

Critical

Ai Agent Control Hijack

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

scripts/setup.mjsView on unpkg · L7

24Manifest entrypoint (scripts.postinstall) carries capability families absent from dist/build output: environment+network, execution+network L24: L25: import { spawnSync } from "node:child_process"; L26: import * as fs from "node:fs"; ... L39: /** postinstall은 조용해야 한다. 명시 실행(npx)일 때만 수다스러워진다. */ L40: const explicit = process.argv.includes("--explicit") || !process.env.npm[redacted]; L41: const silent = process.env.npm[redacted] === "silent"; ... L58: function resolveConsumerRoot() { L59: const candidate = explicit ? process.cwd() : process.env.INIT_CWD; L60: if (!candidate) return null; ... L88: L89: const manifestPath = path.join(root, "package.json"); L90: if (!fs.existsSync(manifestPath)) return "프로젝트 package.json 없음";

High

Entrypoint Build Divergence

Manifest entrypoint contains risky behavior absent from dist/build output.

scripts/setup.mjsView on unpkg · L24

•matchType = normalized_sha256 matchedPackage = @impakers/debug@1.9.7 matchedPath = scripts/setup.mjs matchedIdentity = npm:QGltcGFrZXJzL2RlYnVn:1.9.7 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.

scripts/setup.mjsView on unpkg

dist/debug-widget-DVG4YNS5.jsView file

•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.

dist/debug-widget-DVG4YNS5.jsView on unpkg

Source & flagged code

5 flagged

package.jsonView file

Lines 99-139json

99 "require": {

100 "types": "./dist/telemetry-next.d.ts",

101 "default": "./dist/telemetry-next.js"

102 }

103 }

104 },

105 "bin": {

106 "impakers-debug-setup": "./scripts/setup.mjs"

107 },

108 "files": [

109 "dist",

110 "scripts/setup.mjs",

111 "templates",

112 "CHANGELOG.md"

113 ],

114 "scripts": {

115 "build": "tsup",

116 "watch": "tsup --watch",

117 "dev": "pnpm build && pnpm watch",

118 "prepublishOnly": "pnpm build",

119 "postinstall": "node ./scripts/setup.mjs",

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg · L119

120 "test": "vitest run",

121 "test:watch": "vitest"

122 },

123 "peerDependencies": {

124 "eslint": ">=8.57.0",

125 "next": ">=13.4.0",

126 "react": ">=18.0.0",

127 "react-dom": ">=18.0.0"

128 },

129 "peerDependenciesMeta": {

130 "eslint": {

131 "optional": true

132 },

133 "next": {

134 "optional": true

135 },

136 "react": {

137 "optional": true

138 },

139 "react-dom": {

scripts/setup.mjsView file

Lines 1-44javascript

1#!/usr/bin/env node

2// =============================================================================

3// @impakers/debug — 설치 시 자동 셋업

4// =============================================================================

5//

6// `npm i @impakers/debug` 하면 postinstall로 이 스크립트가 돌면서 두 가지를 한다.

7//

8// 1. .claude/skills/impakers-ui-mapping/SKILL.md 설치

9// 에이전트가 UI 코드를 짤 때 파일 매핑 규칙을 알고 있게 만든다.

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/debug-widget-DVG4YNS5.jsView file

•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.

dist/debug-widget-DVG4YNS5.jsView on unpkg