nihaisha

Warn

Audited by Socket on Aug 10, 2026

1 alert found:

Anomaly
AnomalyLOW
nihaisha_kg/pdf_vector.py

No clear evidence of intentional malware (no exec/eval, no shelling out, no explicit theft/persistence). However, there are two notable security risks: (1) a very unusual multi-statement SQL template in _read_meta_keys_from_connection() that includes DROP/CREATE statements (potential sabotage/destructive behavior if executed), and (2) optional transmission of embedding inputs (user/PDF text) to a third-party API (SiliconFlow) via HTTPS, which is a data-governance/exfiltration concern. Additional red flags include a possibly malformed/garbled conn.execute fragment in build_faiss_vector_index() (could be transcription or could indicate tampering/incompleteness) and reliance on faiss.read_index() loading files without cryptographic integrity checks.

Confidence: 58%Severity: 60%
Audit Metadata
Analyzed At
Aug 10, 2026, 03:28 PM
Package URL
pkg:socket/skills-sh/juneyaooo%2Fnihaisha-nishi-tcm%2Fnihaisha%2F@a575fd8c86e9d7ac177f6a2bf470aca5906bad1f
Security Audit — socket — nihaisha