ClawHub

SQL

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only SQL guide; its database commands can be dangerous if copied into production, but they are disclosed and aligned with the skill’s purpose.

Install if you want SQL reference material. Before allowing an agent to run any example against a real database, verify the database target, confirm backups, test restores in staging or a new database, and require explicit approval for DROP, ALTER, RESTORE, DELETE, or session-termination commands.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The PostgreSQL restore example uses `pg_restore --clean --if-exists`, which explicitly drops existing objects before restoring, but the surrounding text does not clearly warn that this can destroy live data. In an operations skill, readers may copy commands directly into production environments, so omission of an overwrite/data-loss warning creates a realistic risk of accidental destructive use.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The SQL Server restore example shows a direct `RESTORE DATABASE mydb` command without warning that restoring over an existing database can replace or disrupt current data and service availability. Because this document is operational guidance, the lack of caution increases the chance of accidental destructive execution by administrators.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The query terminating example uses `pg_terminate_backend` on active sessions older than one hour, but does not warn that this can interrupt legitimate workloads, abort transactions, and affect application stability. In a database operations context, such commands are powerful administrative actions and need explicit guidance to avoid indiscriminate use.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal