ClawHub
Back to skill

Security audit

postgresql-skill

Security checks across malware telemetry and agentic risk

Overview

This is a real PostgreSQL utility, but one schema command can automatically print sample rows from every public table and the skill can modify database records.

Install only for databases where the agent is allowed to inspect table contents and make row-level changes. Prefer a read-only or limited database user unless writes are needed, avoid running schema-summary on sensitive production databases, and require explicit confirmation before INSERT, UPDATE, or DELETE operations.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The schema-summary feature goes beyond structure discovery and actively queries up to three rows from every table, which can expose sensitive production data such as credentials, tokens, PII, or business records. Because this command is broad and automatic across all public tables, a user invoking what sounds like a metadata operation could unintentionally dump large amounts of confidential information.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The schema-summary command retrieves and prints sample data from every table without any warning, confirmation, or sensitivity filtering. In a database tooling skill, this context makes the issue more dangerous because operators may expect schema inspection, not data exfiltration, and the command can expose confidential data at scale in a single invocation.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill accepts arbitrary SQL so long as it begins with SELECT/INSERT/UPDATE/DELETE and passes simplistic pattern checks, allowing real data modification. In this skill context, arbitrary query execution is expected, but the lack of strong user-facing warning, read-only default, or transaction safety increases the chance of accidental or harmful writes, especially since UPDATE/DELETE protections are easily satisfied with a broad WHERE clause.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.