Security audit

360disk-cli

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent cloud-drive CLI integration, but it gives an agent broad cloud-file control, credential handling, and persistent local-to-cloud backup capability without enough safety scoping.

Review before installing. Use a least-privilege API key, avoid pasting real keys into command lines or shared logs, and require the agent to show exact local and cloud paths before deletes, overwrites, shares, restores, backups, or syncs. Enable auto-backup only for a narrow non-sensitive directory, then verify status and know how to disable it.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (10)

Context-Inappropriate Capability

Medium

Confidence: 89% confidence
Finding: The skill exposes a background daemon that continuously watches a local directory and uploads changes to the cloud, which materially increases the risk of unintended data exfiltration and persistent monitoring. In an agent/tooling context, long-lived background behavior is more dangerous than one-shot file operations because it can continue after the initiating session ends.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The README instructs users to pass an API key directly on the command line (`--api-key <YOUR_API_KEY>`) without any warning about secret handling. Command-line arguments can be exposed through shell history, process listings, CI logs, or copied terminal transcripts, which creates a realistic risk of credential disclosure.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill explicitly instructs the agent to directly execute the most suitable command and exposes destructive operations such as delete, clear-dir, overwrite-style config writes, move/rename, and backup/restore flows, but it provides no requirement for confirmation, dry-run, or user warning before irreversible actions. In an agent setting, this materially increases the chance of accidental destructive execution from ambiguous prompts or misinterpretation, especially because the tool operates on remote cloud data rather than a disposable local sandbox.

Missing User Warnings

Medium

Confidence: 83% confidence
Finding: The authentication section directs use of --api-key, API_KEY, and persistent local storage in ~/.360disk/config.json, but it omits safeguards about secret handling, shell history leakage, process-list exposure, and protection of stored credentials. In an agent or CI context, this can lead to accidental disclosure of API keys through command logs, terminal history, screenshots, or insecure filesystem permissions.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The documentation states that `auth login` stores the API key in `~/.360disk/config.json` but gives no warning about local credential exposure, file permissions, or secure alternatives. Plain local storage of API credentials can lead to account compromise if the workstation, home directory, backups, or logs are accessible to other users or malware.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The `file share` command creates or propagates file-sharing access without any documented warning about exposure scope, downstream redistribution, or sensitive-data handling. In a cloud-storage tool, link-based sharing can easily result in unintentional disclosure if users assume links are private or revocable by default.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: `file url` retrieves direct download links, and the docs do not warn that anyone possessing such a link may gain access depending on backend controls. In practice, copied links often leak through terminals, logs, chat history, or CI output, expanding access beyond the intended recipient.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: `file config` allows direct key edits and whole-file overwrite of cloud-hosted configuration, but the docs do not emphasize the risk of breaking applications, overwriting valid config, or propagating insecure settings. In CI or scripted use, a malformed or misplaced write could silently damage production configuration.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: `file save --url` pulls content from an arbitrary remote URL into cloud storage without warning that this discloses request metadata to the remote server and may ingest malicious or untrusted content. In agent and automation settings, this can be abused as a lightweight SSRF-like fetch primitive or as a means to import hostile payloads into a trusted storage environment.

Missing User Warnings

High

Confidence: 95% confidence
Finding: `claw-auto-backup enable` launches a persistent background process that continuously uploads local file changes, but the docs do not prominently warn about ongoing collection, privacy exposure, or accidental syncing of secrets. Because the process survives the shell session, users may underestimate its persistence and continue generating sensitive data that is silently sent off-host.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.