Chrome Web Store Verified

MCP-B Agent

by [email protected] · 1.0K users · 3.0 rating
926ef15b-442c-509e-9165-570d617e803d | v1.1.0.243
0/ 100
MINIMAL risk
-58 since v1.0.2
58 → 0 · false positives removed
Analyst verdict
No high-risk signal observed

Based on the RiskyPlugins AI security review of the observed evidence.

Analysis record

Analysed
1 months ago
Version
v1.1.0.243
Artifact
SHA256 490…A6D
Source
Findings (non-IoC)

No Findings

All security checks passed

Publisher Evidence

Limited evidence

[email protected]

Publisher identity, store signals, distribution reach, and warning signals used for context. Treat this as supporting evidence, not a clean bill of health.

Chrome does not expose the same publisher verification data as IDE stores, so this score is deliberately conservative.

37
Noisy-finding weight
x1.00
Publisher domain
mcp-b.ai
Observed
Store verification signal
Limited signal
Limited
Extension portfolio
2
Portfolio

12 evidence rows available.

No Threats Detected

This extension passed all security checks

AI Security Report

AI Security Review

Evidence context: threat category none; evidence quality weak.

Security Analysis: MCP-B Extension

Executive Summary

This extension shows no evidence of malicious behavior. All 2,128 findings stem from documented false-positive patterns in the CVEQ analysis pipeline.

Finding Analysis

IoC Findings (1,988 total)

Every IoC finding is a false positive from the XIOC extractor misreading JavaScript property access chains as domain names. For example, XIOC-DOMAIN-r.content.map from extracted_from_files is not a domain—it is the JavaScript expression r.content.map, a standard array method call. Similarly, XIOC-DOMAIN-a.attachments.map and XIOC-DOMAIN-t.message.parts.map follow the same pattern of nested property access (x.y.z.map). These cannot be valid domains and represent the well-documented XIOC garbage pattern where property chains are extracted as if they were hostnames.

Obfuscation Findings (3 total)

The obfuscation detections OBFUSCATION-LARGE_WASM_FILE-esbuild.wasm-0 and OBFUSCATION-LARGE_WASM_FILE-assets/esbuild-BxANhyB7.wasm-0 are false positives. esbuild is a legitimate JavaScript bundler, and its WASM binary is expected in modern web applications. This is not obfuscation—it is a standard build tool artifact.

Network Findings (3 total)

The network findings NET-FETCH-chunks/_commonjsHelpers-DKZ2hJOu.js-1 and NET-FETCH-chunks/index.esm-lnPDxfWZ.js-1 reference bundled chunk files in the chunks/ directory. This is typical webpack/esbuild output and does not indicate suspicious external communication.

Manifest Analysis

The MANIFEST-SENSITIVE-PERM-TABS finding in manifest.json is expected behavior. An AI-powered browser assistant that interacts with web pages requires tabs permission to function as described.

Code-Smell (133 findings)

All 133 code-smell findings are classified as low severity and represent noise from generic YARA rules matching basic JavaScript patterns. These do not indicate malicious behavior.

Counterargument

A skeptic might argue the extension is risky because the developer uses an email address ([email protected]) rather than a verified company name, and the extension requests tabs permission. However, the absence of any malware signatures, the demonstrably false nature of all IoC findings (property chains like r.content.map cannot be domains), and the legitimate nature of the esbuild WASM files override these concerns. The tabs permission aligns with the extension's stated purpose as a browser assistant. Without actual suspicious domains or malicious code patterns, anonymous publishing alone does not constitute a security risk.

Conclusion

This extension is benign. The high finding count is entirely driven by known false-positive patterns in the CVEQ analysis pipeline.

Key Reasons

  • IoC findings are XIOC extractor garbage (property access chains misread as domains)
  • Obfuscation findings are esbuild WASM files (legitimate build tool)
  • Zero malware signatures or actual malicious code indicators
  • Network findings reference bundled chunk files, not suspicious external domains

False Positive Considerations

  • XIOC property access chain false positives (r.content.map, a.attachments.map, etc.)
  • esbuild WASM files flagged as obfuscation
  • Code-smell findings (133 low-severity matches)
  • Bundled chunk files triggering network findings

Reviewed 2026-04-29; recommended action: suppress false positive; model confidence 85%.

About This Extension

Give your AI agent a browser. Think Browser Host helps you get work done across the websites, services, and files you already use. Ask it to research across tabs, organize your browser, work with connected apps, create files, or repeat a task on a schedule. You can: - research and take action across multiple tabs; - organize tabs, bookmarks, history, and downloads; - work with Gmail, Google Calendar, Google Drive, and other connected services; - create reusable skills, scripts, conversations, and files; - choose local folders for a task; and - schedule recurring routines. You stay in control. Browser actions are visible, sensitive changes follow your approval settings, and sign-in or other human-only steps wait for you in the affected tab. Turn one-off work into reusable skills and routines, then run them again whenever you need them. Sign in with ChatGPT to get started. Connected services and local folders are optional—you choose what the agent can use.

Frequently Asked Questions