registry / AI-agent capability abuse / @melandlabs/opencontext / 0.11.0

@melandlabs/opencontext@0.11.0

OpenContext — single-package facade that bundles the contracts, memory-store, retrieval, loop engine, and agent runtime. Install once, get every capability.

AI Security Review

scanned 1mo ago · by lpm-firewall-ai

No concrete install-time malware chain or credential exfiltration was confirmed. The shipped agent/browser and command-execution capabilities remain a real, high-risk user-invoked surface.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

A host application or explicit user action invokes the bundled agent and native sandbox capabilities.

Impact

An agent prompt or untrusted task could access authenticated browser state or execute commands with available user credentials.

Mechanism

Agent-directed browser-session bypass and child-process execution with inherited environment variables.

AI rationale

No concrete install-time malware chain or credential exfiltration was confirmed. The shipped agent/browser and command-execution capabilities remain a real, high-risk user-invoked surface.

Evidence

dist/cli/opencontext.js

Decision evidence

public snapshot

AI called this Suspicious at 86.0% confidence as Dangerous Capability with medium false-positive risk.

Evidence for warning

The published CLI embeds agent instructions to use an active browser session to bypass authentication and anti-bot controls.
Its native sandbox runner starts arbitrary commands with the parent environment merged into the child environment.
The bundled agent runtime enumerates cloud API-key variables for propagation.

Evidence against

package.json has no preinstall, install, postinstall, or self-dependency.
The exposed CLI routes are user-invoked commands; no install-time execution was found.
OpenRouter requests require an explicitly selected feature and a caller-supplied environment key.

Behavioral surface

Source

ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystemNetworkShell

Supply chain

HighEntropyStringsUrlStrings

ManifestNo manifest risk signals triggered.

scanned 2 file(s), 534 KB of source, external domains: 127.0.0.1, github.com, jimmy.warting.se, openrouter.ai, sheetjs.com

Source & flagged code

15 flagged

dist/cli/opencontext.jsView file

17import 'pg'; L18: import { execFile, execFileSync, spawn, execSync, spawnSync, exec } from 'child_process'; L19: import * as z4 from 'zod/v4'; ... L36: import { createRequire } from 'module'; L37: import Xoe from 'http'; L38: import Qoe from 'https'; ... L89: function getTauriDataDir() { L90: if (process.env.TAURI_DATA_DIR) { L91: return process.env.TAURI_DATA_DIR; ... L93: if (process.env.NODE_ENV === "development") { L94: return process.platform === "win32" ? `${process.env.APPDATA}/opencontext` : process.platform === "darwin" ? `${process.env.HOME}/Library/Application Support/opencontext` : `${proc... L95: }

Critical

Credential Exfiltration

Source appears to send environment or credential material to an external endpoint.

dist/cli/opencontext.jsView on unpkg · L17

17Trigger-reachable chain: manifest.bin -> dist/cli/opencontext.js L17: import 'pg'; L18: import { execFile, execFileSync, spawn, execSync, spawnSync, exec } from 'child_process'; L19: import * as z4 from 'zod/v4'; ... L36: import { createRequire } from 'module'; L37: import Xoe from 'http'; L38: import Qoe from 'https'; ... L89: function getTauriDataDir() { L90: if (process.env.TAURI_DATA_DIR) { L91: return process.env.TAURI_DATA_DIR; ... L93: if (process.env.NODE_ENV === "development") { L94: return process.platform === "win32" ? `${process.env.APPDATA}/opencontext` : process.platform === "darwin" ? `${process.env.HOME}/Library/Application Support/opencontext` : `${proc... L95: }

Critical

Trigger Reachable Dangerous Capability

A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.

dist/cli/opencontext.jsView on unpkg · L17

17import 'pg'; L18: import { execFile, execFileSync, spawn, execSync, spawnSync, exec } from 'child_process'; L19: import * as z4 from 'zod/v4';

High

Child Process

Package source references child process execution.

dist/cli/opencontext.jsView on unpkg · L17

201348var jx = {}; L201349: Dt(jx, { USE_API_CONTEXT_MANAGEMENT: () => Rle, INK_SCREEN_READER: () => Jse, FORCE_VCR: () => Cle, FORCE_PROMPT_CACHING_5M: () => Ale, FORCE_CODE_TERMINAL: () => Tle, FORCE_AUTOUP... L201350: var Vse = p.triBool();

High

Shell

Package source references shell execution.

dist/cli/opencontext.jsView on unpkg · L201348

205098var ST = O(() => x({ source: Vp().describe("Where to fetch the marketplace from"), installLocation: g().optional().describe("Local cache path where marketplace manifest is stored (... L205099: var vh = O(() => x({ serverName: g().regex(/^[a-zA-Z0-9_-]+$/, "Server name can only contain letters, numbers, hyphens, and underscores").optional().describe("Name of the MCP serve... L205100: var wh = O(() => x({ serverName: g().min(1, "Server name must be non-empty").refine((e) => e.trim().length > 0, { message: "Server name must not be whitespace-only" }).refine((e) =... ... L205105: function HG(e) { L205106: return x({ $schema: g().optional().describe("JSON Schema reference for Claude Code settings"), apiKeyHelper: g().optional().describe("Path to a script that outputs authentication v..

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/cli/opencontext.jsView on unpkg · L205098

•Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration. dist/cli/opencontext.js: import { readFile as readFile$1, stat as stat$1, mkdir, writeFile as writeFile$1, open as open$1, chmod, link, symlink, appendFile, copyFile, readlink, realpath, rename, rm, rmdir,... import { createInterface } from 'readline'; async function writeFile(path6, data, options) { await writeFile(path6, data); // Marks visible continuation messages created after an authorization card is authorizationContinuation: z$1.boolean().optional(), async function writeFileAtomic(filePath, content) { await nodeFsPromises.writeFile(tempPath, content, {

High

Credential Redirect Persistence

Manifest-reachable source sends a prompted API credential to a fixed unofficial gateway and persists the redirection.

dist/cli/opencontext.jsView on unpkg

•Manifest-reachable source resolves another installed package, overwrites its runtime code, and injects package-defined remote behavior. dist/cli/opencontext.js: import { readFile as readFile$1, stat as stat$1, mkdir, writeFile as writeFile$1, open as open$1, chmod, link, symlink, appendFile, copyFile, readlink, realpath, rename, rm, rmdir,... "follow up" // ../../node_modules/clsx/dist/clsx.mjs "../../node_modules/clsx/dist/clsx.mjs"() { // ../../node_modules/@tauri-apps/api/external/tslib/tslib.es6.js "../../node_modules/@tauri-apps/api/external/tslib/tslib.es6.js"() { "../../node_modules/@tauri-apps/api/core.js"() { // ../../node_modules/@tauri-apps/api/path.js

High

Entrypoint Foreign Package Code Overwrite

Manifest-reachable source overwrites another installed package with package-defined remote behavior.

dist/cli/opencontext.jsView on unpkg

High

Command Output Exfiltration

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

dist/cli/opencontext.jsView on unpkg · L17

High

Sandbox Evasion Gated Capability

Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.

dist/cli/opencontext.jsView on unpkg · L17

17import 'pg'; L18: import { execFile, execFileSync, spawn, execSync, spawnSync, exec } from 'child_process'; L19: import * as z4 from 'zod/v4'; ... L7693: function writeFileSync2(filePath, content) { L7694: we.writeFileSync(filePath, content); L7695: } ... L12174: // ../../node_modules/langsmith/dist/_openapi_client/client.js L12175: var __classPrivateFieldSet4, __classPrivateFieldGet4, _Langsmith_instances, _a, _Langsmith_encoder, _Langsmith_baseURLOverridden, Langsmith; L12176: var init_client = __esm({ ... L135872: availableTools = void 0, L135873: message = `Model tried to call unavailable tool '${toolName}'. ${availableTools === void 0 ? "No tools are available." : `Available tools: ${availableTools.join(", ")}.`}` L135874: }) {

High

Remote Agent Bridge

Source exposes local file and command tools to a remote model endpoint.

dist/cli/opencontext.jsView on unpkg · L17

17Trigger-reachable credential exfiltration chain: manifest.bin -> dist/cli/opencontext.js L17: import 'pg'; L18: import { execFile, execFileSync, spawn, execSync, spawnSync, exec } from 'child_process'; L19: import * as z4 from 'zod/v4'; ... L36: import { createRequire } from 'module'; L37: import Xoe from 'http'; L38: import Qoe from 'https'; ... L89: function getTauriDataDir() { L90: if (process.env.TAURI_DATA_DIR) { L91: return process.env.TAURI_DATA_DIR; ... L93: if (process.env.NODE_ENV === "development") { L94: return process.platform === "win32" ? `${process.env.APPDATA}/opencontext` : process.platform === "darwin" ? `${process.env.HOME}/Library/Application Support/opencontext` : `${proc... L95: }

High

Trigger Reachable Credential Exfiltration

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

dist/cli/opencontext.jsView on unpkg · L17

•stage = ast_semantic_analysis; reason = ast_path_work_budget_exceeded; 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/cli/opencontext.jsView on unpkg

10895const isOldNode = typeof process2?.versions?.node === "string" && parseInt(process2.versions.node.split(".")) < 20; L10896: throw new Error("`File` is not defined as a global, which is required for file uploads." + (isOldNode ? " Update to Node 20 LTS or newer, or set `globalThis.File` to `import('node:... L10897: }

Medium

Dynamic Require

Package source references dynamic require/import behavior.

dist/cli/opencontext.jsView on unpkg · L10895

•path = dist/cli/opencontext.js kind = oversized_cli_entrypoint sizeBytes = 8363178 magicHex = [redacted]

Medium

Oversized Cli Entrypoint

Package contains an oversized executable-looking CLI entrypoint.

dist/cli/opencontext.jsView on unpkg

dist/index.jsView file

•path = dist/index.js kind = oversized_source_file sizeBytes = 10083556 magicHex = [redacted]

Medium

Oversized Source File

Package contains source files above the normal full-analysis size ceiling.

dist/index.jsView on unpkg

Source & flagged code

15 flagged · loading source

dist/cli/opencontext.jsView file

Critical

Credential Exfiltration

Source appears to send environment or credential material to an external endpoint.

dist/cli/opencontext.jsView on unpkg · L17

Critical

Trigger Reachable Dangerous Capability

A package entrypoint or install-time lifecycle script reaches a source file with blocking dangerous behavior.

Findings

2 Critical10 High6 Medium4 Low

CriticalCredential Exfiltrationdist/cli/opencontext.js

CriticalTrigger Reachable Dangerous Capabilitydist/cli/opencontext.js

HighChild Processdist/cli/opencontext.js

HighShelldist/cli/opencontext.js

HighSame File Env Network Executiondist/cli/opencontext.js

HighCredential Redirect Persistencedist/cli/opencontext.js

HighEntrypoint Foreign Package Code Overwritedist/cli/opencontext.js

HighCommand Output Exfiltrationdist/cli/opencontext.js

HighSandbox Evasion Gated Capabilitydist/cli/opencontext.js

HighRemote Agent Bridgedist/cli/opencontext.js

HighTrigger Reachable Credential Exfiltration

Affected versions and remediation

This report applies to @melandlabs/opencontext@0.11.0.

See all versions of @melandlabs/opencontext for other recorded verdicts.

Review the evidence and your use of @melandlabs/opencontext@0.11.0 before allowing it. Restrict the permissions described in this report.
Choose an independently verified alternative or release. This report does not establish that other versions are safe.

Evidence last updated: 2026-09-12 10:15:09.410Z (UTC).

Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Runtime surface

Behavioral surface

Source & flagged code