VS Code Marketplace Store verified

Paver

by Red Hat · 221 users
2e32b9df-21d6-5029-948e-b756f249a6a5 | Store v0.1.2025021433
76/ 100
HIGH risk
AI review verdict
Review before use

The AI review rates the findings as likely false positive. The risk score (76/100) comes from stored counts that the review could not adjust. A new analysis will update it.

Analysis record

Analysed
11 months ago 1 December 2025
Version
v0.1.2025021433
Artifact
SHA256 dd2ce959…334ef5
Detail
Counts only, rows expired

Detail for this analysis has expired

This analysis ran on 1 December 2025. 33 findings counted, and the counts are kept. The individual rows have aged out of storage. A new analysis restores them.

Detail for this analysis has expired

This analysis ran on 1 December 2025. 457 network indicators counted, and the counts are kept. The individual rows have aged out of storage. A new analysis restores them.

Publisher Evidence

Low

Red Hat

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

86
Publisher domain
redhat.com
Observed
Store verification signal
Verified publisher
Verified
Extension portfolio
64
Portfolio

11 evidence rows available.

Detail for this analysis has expired

This analysis ran on 1 December 2025. 490 findings counted, and the counts are kept. The individual rows have aged out of storage. A new analysis restores them.

Stored counts by category

457
Network indicators
29
Code patterns
29
Malware signatures
2
Dependencies
2
Network calls

AI Security Report

AI Security Review

Evidence context: threat category none; evidence quality moderate.

The Paver extension from Red Hat presents 490 total findings, but these are overwhelmingly false positives from the XIOC extractor rather than actual security risks.

Filesystem and Process Access Justification: The extension's stated purpose is to "Use IBM Granite LLM as your Code Assistant." The single network finding NET-FETCH-extension/dist/extension.js-9 shows a fetch call in the main extension file, which is expected behavior for an LLM assistant that must communicate with IBM Granite's API. No postinstall payload execution, shell command execution, or suspicious process spawning was detected. The extension has no findings in categories that would indicate excessive capability beyond its stated purpose.

Credential-Access Findings: The evidence bundle shows zero findings in the "secret" category. None of the IoC findings target actual secrets like .env files, .ssh directories, cloud credentials, or VS Code's secret storage. The 29 code-smell findings are low-severity noise per CVEQ's documented false positive patterns and do not indicate credential theft.

XIOC False Positive Pattern: Of the 457 IoC findings, 9 shown in the evidence are classic XIOC extractor false positives: s.call, t.style, p.call, m.now, d.now, l.call, id-z.id, and t.is are JavaScript property access chains (e.g., s.call(), performance.now()) misidentified as domains. The finding contributing.md is a filename, not a domain. Only one legitimate domain appears: docs.continue.dev, which is Continue.dev's documentation site. This pattern matches CVEQ's documented IoC extractor garbage.

Strongest Counterargument: The low user count (221) and recent version number (0.1.2025021433) could suggest a new or unverified extension. However, the developer is Red Hat, a legitimate enterprise software company, and the extension's purpose (IBM Granite LLM integration) is clear and legitimate. The high finding count is explained by XIOC false positives, not malicious code. No evidence supports supply chain poisoning, credential theft, or exfiltration.

The verdict is likely_false_positive because all findings are explainable as XIOC noise or expected IDE extension behavior for an LLM assistant.

Key Reasons

  • Developer is Red Hat (legitimate enterprise publisher)
  • IoC findings are XIOC false positives (property access chains, filenames)
  • Network fetch in extension.js is expected for LLM assistant
  • Zero credential/secret findings detected
  • No postinstall payload or suspicious process execution

False Positive Considerations

  • XIOC property access chains misidentified as domains (s.call, t.style, m.now, etc.)
  • Filename contributing.md extracted as domain
  • Code-smell findings (29 total) are documented noise
  • High IoC count (457) from minified/bundled JavaScript

Reviewed 2026-04-29 · version 0.1.2025021433; recommended action: suppress false positive; model confidence 85%.

Source Code Viewer

Source Code Not Available

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

About This Extension

Use IBM Granite LLM as your Code Assistant

Frequently Asked Questions