Security audit

analytics

Security checks across malware telemetry and agentic risk

Overview

This is a non-executable analytics guidance skill, but users should apply privacy and consent controls before copying its tracking examples.

Install only if you want analytics implementation guidance. Before using its GA4, GTM, or Facebook Pixel examples, review consent requirements, avoid sending PII or raw user identifiers, use pseudonymous IDs where needed, and ensure third-party tags fire only under the consent and privacy rules that apply to your users.

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 (5)

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The eval explicitly rewards the skill for triggering on casual phrasing ('our tracking seems broken') without requiring a tighter boundary check beyond general analytics context. This can cause over-broad skill activation, making the agent respond with specialized analytics guidance in ambiguous situations and increasing the chance of misrouting user requests away from more appropriate skills or safer clarification flows.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The reference explicitly recommends collecting persistent identifiers such as user_id and account_id along with referrer and campaign context, but provides no privacy guidance, consent considerations, data minimization advice, or warnings about handling personal data. In an analytics implementation skill, this omission can lead downstream users to instrument tracking in ways that violate privacy expectations, internal policy, or legal requirements, especially when combined with cross-session/user profiling.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The reference includes examples that set `user_id` and user properties such as `user_type` and `plan_name` without an adjacent warning about consent, privacy obligations, or avoiding direct identifiers. In an analytics implementation skill, readers may copy these snippets verbatim, which can lead to privacy noncompliance, over-collection, or transmission of personal data to GA4 in ways that violate policy or regulation.

Missing User Warnings

Medium

Confidence: 85% confidence
Finding: The reference explicitly shows adding user-related fields such as userId, loggedIn status, and userType to the data layer for downstream analytics use, but does not warn about privacy, consent, minimization, or restrictions on sending identifiers to analytics/ads platforms. In an analytics implementation skill, this omission is materially risky because practitioners may copy the pattern directly and unintentionally transmit personal or pseudonymous identifiers in ways that violate platform policies or privacy requirements.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The Facebook Pixel examples load a third-party tracker on all pages and send form-related event data, yet the document provides no warning that this shares browsing and interaction data with Meta or that consent may be required before firing such tags. Because this file is an implementation guide, users are likely to reproduce the snippet verbatim, increasing the chance of unauthorized third-party data disclosure and noncompliant tracking.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.