Security audit

agentdrive-cli

Security checks across malware telemetry and agentic risk

Overview

This is a coherent cloud-drive CLI skill, but it gives an agent broad file-management, local backup/restore, credential, and persistent auto-sync authority without enough built-in confirmation or scoping guidance.

Install only if you are comfortable letting an agent manage your AgentDrive files and any local folders you explicitly point it at. Prefer an environment variable or protected config over putting API keys in commands, pin the npm package version for repeatability, and require explicit confirmation before delete, clear-dir, overwrite/replace, restore, logout, backup, or auto-backup enable commands.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (8)

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The documented CLI exposes capabilities beyond ordinary remote cloud-file manipulation: recursive local backup, restore to the local filesystem, and persistent background sync. In an agent skill context, these features materially expand the trust boundary because the tool can read local data, write to local paths, and keep operating after the initiating session, increasing the risk of unintended data exfiltration or persistence.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The automatic backup command starts a background daemon that continuously watches a local directory and synchronizes changes to the cloud after the shell session ends. In an agent environment, that persistence is especially dangerous because it can create ongoing monitoring and exfiltration behavior not obvious from a one-shot cloud storage CLI description.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The README instructs users to pass an API key directly on the command line without warning about credential exposure. Command-line secrets can leak through shell history, process listings, CI logs, or telemetry, which is especially risky for an agent skill intended for automation and scripting.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill explicitly supports destructive operations such as delete, clear-dir, move/replace, and logout/clear local config, while also instructing the agent to 'directly execute the most appropriate command' without requiring confirmation for destructive actions. In an agent setting, this increases the risk of accidental or prompt-induced irreversible cloud data loss because a single mistaken interpretation can trigger deletion or overwrite immediately.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The authentication flow directs the agent to check login state each session and, if needed, run login using an API key or local config sources, but it does not warn about handling secrets or reading credentials from local files such as ~/.agentdrive/config.json. In practice, this can lead to unnecessary secret exposure in prompts, shell history, logs, CI output, or unintended access to locally stored credentials without clear user consent.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The backup documentation explains how to upload local files or directories to the cloud but does not clearly foreground the privacy consequence that local content will leave the machine and be stored remotely. For an agent-operated tool, omission of that warning can cause users to authorize uploads of sensitive local data without informed consent.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The restore command writes cloud content onto the local filesystem, but the docs do not clearly warn about overwrite, merge, or contamination risks in the target directory. In automated or agent-driven use, that can lead to unexpected local file creation and damage to existing data layouts.

Missing User Warnings

High

Confidence: 94% confidence
Finding: The auto-backup section mentions a background daemon but does not prominently warn that enabling it continuously monitors a local directory and uploads subsequent changes to the cloud. That omission is severe in a skill context because users may perceive a transient command while actually authorizing long-lived surveillance and exfiltration behavior.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.