@ripla/godd-mcp

AI Security Review

Evidence context: threat category none; evidence quality weak.

This package presents a critical risk score of 100 based on volume alone, with 2,030 IoC findings and 237 malware signatures. However, this pattern is a textbook false-positive profile for MCP servers containing bundled JavaScript dependencies. The scanner has detected zero instances of tool poisoning, zero secret exposure, and only three generic network findings. The "malware" signatures are likely legacy YARA rules (e.g., Surtr, Bolonyokte) firing on generic configuration strings or minified code within the bundle. The high IoC count consists primarily of property access chains (b.call, h.next), IPv6 fragments, and CDNs misidentified as network threats. There is no evidence of environment variable exfiltration, no tool result interception, and no mismatch between the package's stated purpose (encrypted prompt distribution) and the lack of specific threat indicators. The 'CRITICAL' rating appears to be a score inflation artifact, not a reflection of actual malicious behavior.

Key Reasons

• Zero tool-poisoning findings despite 2,500+ total matches
• No credential-access or secret-exposure findings
• IoC and signature counts consistent with bundled dependencies causing false positives
• Absence of suspicious network domains (only 3 generic findings)

False Positive Considerations

• Bundled dependencies (dist/ files)
• YARA code-smell rules (generic malware families)
• IoC extractor garbage (property chains, hex strings)

Reviewed 2026-04-12; recommended action: suppress false positive; model confidence 85%.