From the RiskyPlugins AI security review of the observed evidence. The review covers version 3.4.1, not the analysed version 4.5.0.
Analysis record
- Analysed
- 2 months ago 26 August 2026
- Version
- v4.5.0
- Artifact
- SHA256 4a55fda5…c0bf59
- Detail
- Full list
Evidence ledger
Ranked by severity · findings with a source location link to the code viewer
Finding Categories
MCP Server Analysis
MCP servers expose tools and resources to AI assistants. Unlike browser extensions, they run as standalone processes with direct system access. Tool definitions are analyzed for prompt injection, data exfiltration, and tool poisoning patterns.
AI Security Report
AI Security Review
Evidence context: threat category unknown malware; evidence quality weak.
This package @kernlang/review-mcp describes itself as an "MCP server security scanner" that performs static analysis on Model Context Protocol implementations. The findings require careful interpretation given the package's stated purpose.
Tool-Poisoning Findings: The findings summary reports 8 tool-poisoning findings, but the detailed evidence does not include the specific finding titles or file paths for these. Given this package is a security scanner that likely defines its own MCP tools for analysis, these findings are potentially false positives from legitimate tool definitions. However, without the actual finding details, I cannot confirm whether these are hidden AI directives or benign tool descriptions. The absence of specific tool-poisoning evidence in the findings_by_category prevents definitive assessment.
Credential/Secret Access: The findings summary indicates 7 secret findings, but these are not included in the provided evidence. This is a critical gap. If these secret findings target sensitive paths like .ssh/, .aws/credentials, or .kube/config, this would indicate credential harvesting. If they are generic process.env reads for configuration, they are normal for MCP servers. The evidence bundle does not show the secret finding details.
Network Activity: Nine network findings appear in src/rule-runner.ts (lines 83, 150), src/rules/checks/mcp12-resource-exhaustion.ts (line 61), and test fixtures. These fetch calls are consistent with a security scanner making API requests. Importantly, there are 0 IoC findings—no suspicious external domains were detected. This is a positive signal indicating the network calls likely target documented endpoints rather than exfiltration servers.
Strongest Counterargument: The package has an unknown publisher (cukasn), zero users, and 15 critical-severity findings. This combination is concerning for an MCP server. However, the critical findings are not detailed in the evidence, and the 0 IoC count suggests no exfiltration architecture. The description as a "security scanner" provides a legitimate explanation for tool definitions triggering tool-poisoning rules.
Conclusion: Manual code review is required to examine the 8 tool-poisoning findings and 7 secret findings directly. The verdict cannot be confirmed without seeing whether tool descriptions contain hidden AI directives and whether secret access targets sensitive credential paths.
Key Reasons
- 8 tool-poisoning findings require manual review to distinguish directives from tool definitions
- 7 secret findings not detailed in evidence—need to verify credential scope
- 0 IoC findings indicate no exfiltration domains detected
- Unknown publisher with zero users increases suspicion baseline
- Package description as security scanner provides legitimate context for findings
False Positive Considerations
- Tool-poisoning findings likely from legitimate MCP tool definitions in security scanner code
- Network findings are standard fetch calls without suspicious domain IoCs
- Bundled dist/ files create duplicate findings from source code
Reviewed 2026-05-03 · version 3.4.1; recommended action: reanalyze; model confidence 65%.
MCP version history
Risk trend by version
21 analyzed versions. Each point is the latest successful scan for that version; failed zero-score scans are hidden. Dates: First seen by Risky Plugins.
Pick any point on the chart, then open the Source tab to read that version's code.
Source Code Not Available
Source code is not available for this version of the extension.
About This Extension
Frequently Asked Questions
Similar Extensions
Related extensions from the same publisher or marketplace