Sniplink

Security checks across malware telemetry and agentic risk

Overview

Sniplink appears to be a legitimate URL saver, but it is published with a hardcoded personal Google Drive Obsidian path and writes persistent notes there after approval.

Install only if you are comfortable with the skill fetching URLs you provide and saving approved records into an Obsidian vault. Before use, replace the published Google Drive path with your own chosen folder, confirm each write destination, and avoid submitting private or sensitive URLs unless you intend their metadata to be fetched and stored.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: The skill instructs the agent to invoke shell commands and contact third-party services such as curl to api.fxtwitter.com and gh CLI operations. That expands the trust boundary well beyond simple URL saving and can leak user-provided URLs, tweet references, or derived targets to external systems without a narrowly scoped consent model.

Context-Inappropriate Capability

Low

Confidence: 85% confidence
Finding: The social-media lookup step directs the agent to search for LinkedIn and Twitter profiles for a tool and store those URLs, which is ancillary to the core task of saving a submitted link. This increases data collection and external querying beyond user necessity, creating avoidable privacy and scope-creep risk.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The trigger phrases include common language such as 'save this' and 'remember this,' which can cause the skill to activate in ordinary conversation when the user may not intend file writes, lookups, or database updates. Overbroad activation increases the chance of unintended external fetches and unintended persistence of data.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill describes writing notes into the user's Obsidian vault but does not provide an upfront warning in the trigger or setup that local files will be created or modified. Users may invoke the skill expecting a lightweight save operation without realizing it persists data into a local/cloud-synced filesystem location.

Natural-Language Policy Violations

High

Confidence: 98% confidence
Finding: The skill hard-codes a specific personal Google Drive-backed vault path containing an apparent personal email address and user-specific directory structure. This can cause writes into the wrong person's storage, expose personal path information, and create unsafe assumptions about the runtime environment and access scope.

Ssd 3

Medium

Confidence: 97% confidence
Finding: The documentation exposes a specific cloud-storage vault path in plain text, including personal directory details. Even without direct exploitation, this is sensitive environment information disclosure that can aid targeting, misdirection of writes, or privacy leakage.

VirusTotal

No VirusTotal findings

View on VirusTotal