Security audit

Social Media Toolkit

Security checks across malware telemetry and agentic risk

Overview

The skill is a disclosed social-media automation toolkit, but it enables bulk external actions without enough scoping, confirmation, or safety controls.

Install only if you control the relevant social-platform accounts and are prepared to approve every external action manually. Treat bulk messaging, bulk likes/swipes, relationship changes, webhook handling, database archival, and API-token use as high-impact operations; verify rate limits, consent, platform policy compliance, and command restrictions before use.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Intent-Code Divergence

High

Confidence: 95% confidence
Finding: The documented input/output contract describes an audit/scoring workflow, while the rest of the skill is a social-media automation toolkit with external side effects. This mismatch can cause an agent to invoke the skill under the wrong assumptions, mis-handle parameters, or trust outputs that do not correspond to the actual behavior, increasing the risk of unintended automation or unsafe execution.

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The skill claims command execution is limited to a whitelist, but no whitelist or enforcement rules are actually defined in the instructions. In a skill with exec capability, this creates a false sense of safety and can lead operators or agents to run arbitrary shell commands or unsafe command variants under the assumption that restrictions exist.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The activation guidance uses broad 'Use when' wording without clear boundaries, making it easy for an agent to select this skill in contexts that only loosely match. Because this skill supports write and exec and describes external API actions, ambiguous triggering increases the chance of unreviewed or inappropriate activation.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The markdown promotes bulk swiping, bulk messaging, and relationship-state orchestration affecting third-party accounts and data, but it does not prominently require user confirmation, authorization checks, rate-safety controls, or warnings about external side effects. In context, this is especially dangerous because the skill is explicitly designed for large-scale social operations and multi-agent coordination, amplifying abuse, spam, privacy harm, and policy violations.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: SKILL.md:71