registry / AI-agent capability abuse / fraim / 2.0.341

fraim@2.0.341

FRAIM core CLI and MCP package.

AI Security Review

scanned 3d ago · by lpm-firewall-ai

Explicit CLI setup enables telemetry globally for Claude Code without a separate telemetry consent step, so a warning is warranted for agent configuration capability. The inspected receiver is loopback-bound and handles token and cost metrics; the source does not establish external credential or prompt export.

Static reason

One or more suspicious static signals were detected.; source matched previously finalized malicious package; routed for review; previous stored version diff introduced dangerous source

Trigger

Running `fraim init-project` when Claude Code is detected, or adding Claude Code through `fraim add-ide`.

Impact

Claude Code token and cost metrics may be collected across projects after setup without a separate telemetry consent step in the inspected path.

Mechanism

The CLI adds missing telemetry settings to `~/.claude/settings.json`, directing OTLP metrics to a FRAIM receiver bound to loopback, which stores token and cost snapshots by session.

AI rationale

Evidence

dist/src/cli/setup/claude-code-telemetry.jsdist/src/cli/commands/init-project.jsdist/src/cli/commands/add-ide.jsdist/src/local-mcp-server/otlp-metrics-receiver.jsdist/src/local-mcp-server/stdio-server.js~/.claude/settings.json

Network endpoints2

localhost:4318127.0.0.1

Decision evidence

public snapshot

AI called this Suspicious at 91.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

package.json exposes a CLI and has no lifecycle scripts.
Running `fraim init-project` with Claude Code detected enables telemetry in user-level settings, which applies across projects.
The helper enables Claude telemetry and points metrics to localhost; it adds settings only when the keys are absent.
The receiver extracts token and cost metrics, stores session snapshots, and binds to loopback.
The diagnostic excerpt redacts secret material, and the first-run process runner disables shell execution.

Evidence against

Telemetry setup is triggered by explicit CLI commands, not npm installation.
Inspected telemetry code collects token and cost metrics through a loopback receiver; it does not establish export to an unrelated external recipient.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStrings

Source & flagged code

16 flagged

dist/src/core/utils/git-utils.jsView file

9exports.sanitizeRepoIdentifier = sanitizeRepoIdentifier; L10: const child_process_1 = require("child_process"); L11: function extractLocalFolderLabel(rawPath) {

High

Child Process

Package source references child process execution.

dist/src/core/utils/git-utils.jsView on unpkg · L9

dist/src/cli/doctor/checks/scripts-checks.jsView file

235// Primary probe: py launcher (not subject to App Execution Alias) L236: const pyResult = (0, child_process_1.spawnSync)('py --version', { timeout: 1500, encoding: 'utf8', shell: true, windowsHide: true }); L237: if (pyResult.status === 0) {

High

Shell

Package source references shell execution.

dist/src/cli/doctor/checks/scripts-checks.jsView on unpkg · L235

bin/fraim.jsView file

3try { L4: require('../dist/src/cli/fraim.js'); L5: } catch (error) {

Medium

Dynamic Require

Package source references dynamic require/import behavior.

bin/fraim.jsView on unpkg · L3

dist/src/cli/utils/persist-shell-path.jsView file

10Trigger-reachable persistence chain: manifest.bin -> bin/fraim.js -> dist/src/cli/fraim.js -> dist/src/cli/commands/first-run.js -> dist/src/first-run/session-service.js -> dist/src/cli/utils/persist-shell-path.js L10: const path_1 = __importDefault(require("path")); L11: const child_process_1 = require("child_process"); L12: const script_sync_utils_1 = require("./script-sync-utils"); ... L51: const stanza = `\n${marker}\n${exportLine}\n`; L52: if (process.platform === 'win32') { L53: const psCmd = [redacted](bins); ... L60: // dedicated flag for this call site. L61: if (process.env.FRAIM_TEST_SANDBOX !== '1') { L62: const result = (0, child_process_1.spawnSync)('powershell', ['-NoProfile', '-NonInteractive', '-Command', psCmd], { encoding: 'utf8', timeout: 10000, windowsHide: true }); L63

High

Trigger Reachable Persistence

A manifest entrypoint or package-local install chain reaches persistence behavior.

dist/src/cli/utils/persist-shell-path.jsView on unpkg · L10

10const path_1 = __importDefault(require("path")); L11: const child_process_1 = require("child_process"); L12: const script_sync_utils_1 = require("./script-sync-utils"); ... L51: const stanza = `\n${marker}\n${exportLine}\n`; L52: if (process.platform === 'win32') { L53: const psCmd = [redacted](bins); ... L60: // dedicated flag for this call site. L61: if (process.env.FRAIM_TEST_SANDBOX !== '1') { L62: const result = (0, child_process_1.spawnSync)('powershell', ['-NoProfile', '-NonInteractive', '-Command', psCmd], { encoding: 'utf8', timeout: 10000, windowsHide: true }); L63: if (result.error || result.status !== 0) { L64: throw new Error(`Could not register managed binaries on the User PATH: ${result.error?.message || result.stderr}`); L65: }

Medium

Install Persistence

Source writes installer persistence such as shell profile or service configuration.

dist/src/cli/utils/persist-shell-path.js

dist/src/cli/commands/override.jsView file

Lines 1-30javascript

1"use strict";

2var __importDefault = (this && this.__importDefault) || function (mod) {

3 return (mod && mod.__esModule) ? mod :

dist/src/cli/setup/user-level-sync.jsView file

133try { L134: (0, child_process_1.execSync)('npm install --no-audit --no-fund --no-save --no-package-lock --omit=dev', { L135: cwd: baseDir,

High

Runtime Package Install

Package source invokes a package manager install command at runtime.

dist/src/cli/setup/user-level-sync.jsView on unpkg · L133

dist/src/cli/utils/managed-agent-install.js#virtual:normalized:round1View file

•stage = ast_semantic_analysis; reason = ast_parse_error; limitedFiles = 9

High

Semantic Analysis Limited

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

dist/src/cli/utils/managed-agent-install.js#virtual:normalized:round1View on unpkg

index.jsView file

•matchType = normalized_sha256 matchedPackage = fraim@2.0.183 matchedPath = index.js matchedIdentity = npm:ZnJhaW0:2.0.183 similarity = 1.000 summary = normalized source hash matched finalized malicious source

High