VS Code Marketplace Verified domain

aianalyzer

by AMD · 170 users · 5.0 rating
43f62d2c-5cea-5763-ab81-abb4630b9336 | Store v1.9.20261020
0/ 100
MINIMAL risk
-44 since v1.9.20261001
AI review verdict
No high-risk signal observed

Based on the RiskyPlugins AI security review of the observed evidence. The review covers version 1.9.20261001, not the analysed version 1.9.20261020.

Analysis record

Analysed
1 month ago 23 October 2026
Version
v1.9.20261020
Artifact
SHA256 3ea11e6a…ae5ef2
Detail
Full list

No findings

The analyzer reported nothing for this version.

Publisher Evidence

Limited evidence

AMD

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

37
Publisher domain
amd.com
Observed
Store verification signal
No verified badge
Not verified
Extension portfolio
3
Portfolio

12 evidence rows available.

No findings

The analyzer reported nothing for this version.

AI Security Report

AI Security Review

Evidence context: threat category none; evidence quality moderate.

AMD's AI Analyzer is a Visual Studio Code extension with an empty permission list. The manifest declares no host permissions, no network endpoints and no activation-time process hooks, so the installed package asks for nothing beyond the default access any VS Code extension inherits.

The only findings in this build sit in a single file: extension/dist/webviews/WorkerLayout.js, at lines 33865, 33882, 33961 and 34045, flagged by the NET-SOCKET_IO rule. That rule matches socket-style client code. A webview worker that talks back to the extension host is how a sidebar panel receives streaming results and repaints them, and the path confirms it: dist/webviews is build output, the bundled UI half of the extension. Nothing in the package reads .env, .git/config, ~/.ssh or cloud credential files. Secret findings are absent, malware signature matches are absent, obfuscation findings are absent, dependency findings are absent, and no tool-poisoning hit was recorded.

That file is a minified bundle, which is why the four hits land at such high line numbers in one path rather than spread across source files. Minified webview JavaScript is build output, not intentional hiding, and a socket client bundled into a UI worker cannot move any data without a destination. No outbound domain was captured anywhere in this package, and no endpoint is declared in the manifest. The total finding count is four, all of one type, all in one file, which is the opposite of the spread you would see if an extension were reaching into a workspace and shipping it somewhere.

Filesystem and process access: none detected. There are no child_process spawn findings, no shell execution findings and no file-read findings outside the webview bundle. For an extension whose stated job is analyzing data for AMD hardware, reading workspace files would be justified anyway, since that is the work. The absence of any such finding leaves nothing to weigh against.

The strongest counterargument is the name. "aianalyzer" is generic, the listing sits at 144 users, and generic names on low install counts are the usual shape of a lookalike package. That shape does not hold up here. The publisher is AMD, the version string 1.9.20261001 tracks a real release cadence rather than a one-off upload, and there is no second listing with a near-identical name carrying broader permissions. A copycat would also need somewhere to send data, and this package names no destination. Socket code in a webview worker that reports results to its own extension host is the ordinary wiring for this kind of tool, not intelligence gathering.

Key Reasons

  • All four findings are NET-SOCKET_IO hits confined to extension/dist/webviews/WorkerLayout.js, a bundled webview worker, with no other file affected.
  • The manifest declares no permissions, no host permissions and no network endpoints, so there is no declared capability to abuse.
  • Zero credential, secret, malware-signature, obfuscation, dependency or tool-poisoning findings in the package.
  • No child_process or shell execution findings, and no file-read findings outside the webview bundle.
  • Published by AMD with a version string tracking a real release cadence, which argues against a typosquat or re-upload.

False Positive Considerations

  • NET-SOCKET_IO rule matching socket client code in minified webview build output under dist/
  • High line numbers in extension/dist/webviews/WorkerLayout.js reflect a bundled file, not concealed logic
  • Sidebar-to-extension-host messaging is normal webview wiring for an analyzer panel
  • No outbound endpoints or domains captured, so the socket code has no exfiltration destination in this package

Reviewed 2026-10-11 · version 1.9.20261001; recommended action: no action; model confidence 85%.

VS Code version history

Risk trend by version

9 analyzed versions. Each point is the latest successful scan for that version; failed zero-score scans are hidden. Dates: First seen by Risky Plugins.

Latest
43
Change since first
+2
Change from previous
No change
Versions:
First analyzed version
1.7.0
First seen Jan 23, 2026
Risk range
41 to 51
Across analyzed versions
Latest analyzed version
1.9.20261001
First seen Oct 9, 2026
Selected version
minimal
Version
v1.9.20261020
Analysed 1 month ago
Risk score
0
Findings
0
Change vs previous
-44

Pick any point on the chart, then open the Source tab to read that version's code.

Source Code Viewer

Source Code Not Available

Source code is not available for this version of the extension.

About This Extension

AMD AI Analyzer extension for Visual Studio Code

Frequently Asked Questions