Back to skill

Security audit

DuckDB CLI skills

Security checks across malware telemetry and agentic risk

Overview

This is a DuckDB CLI reference skill whose local data access and file-writing examples match its stated data analysis purpose.

Install this only if you want DuckDB CLI help. Review generated commands before running them, especially file globs, export paths, COPY targets, CREATE or INSERT statements, .output or .once destinations, and .edit/editor use. Prefer read-only mode when inspecting existing databases.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
90% confidence
Finding
The trigger phrases are very broad, including generic terms like "sql", "query", and "data analysis", which can cause the skill to activate for ordinary requests outside a user's intent. Over-broad routing increases the chance that the agent will apply file-reading, file-writing, or database-modifying guidance in the wrong context, expanding the attack surface and creating unsafe tool-selection behavior.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill contains multiple examples that write files and modify databases, such as `COPY ... TO`, `CREATE TABLE`, and `INSERT INTO`, but it does not warn users that these operations are state-changing and may overwrite or create artifacts. In an agent setting, omission of modification warnings can lead to accidental data loss, unintended persistence, or silent alteration of local datasets when the skill is selected automatically.

Missing User Warnings

Low
Confidence
78% confidence
Finding
The documentation includes capabilities that can launch local programs or interact with the filesystem, such as `.edit`, `.read`, `.output`, and editor selection via environment variables, without warning that these actions may execute local software or affect local files. While these are legitimate DuckDB features, omitting safety guidance in an agent skill can normalize risky operations and make unintended local side effects more likely.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.