Security audit

heursistic

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only persuasion skill with no malware behavior, but it broadly teaches pressure-oriented influence tactics without clear ethical limits.

Install only if you intentionally want an agent to use aggressive persuasion heuristics for sales or outreach. Before using it, add clear boundaries: truthful claims only, factual urgency and scarcity, respectful follow-up limits, clear opt-outs, and no coercion, harassment, deception, emotional exploitation, or targeting of vulnerable people.

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 (5)

Vague Triggers

High

Confidence: 97% confidence
Finding: The activation guidance is intentionally expansive, telling the agent to invoke this persuasion skill for essentially any request involving influence or getting someone to act. That broad trigger scope increases the chance the skill will be used in ordinary conversations, emotionally sensitive contexts, or borderline manipulative tasks without meaningful user consent or safety gating.

Natural-Language Policy Violations

High

Confidence: 99% confidence
Finding: The skill description explicitly frames human psychology as 'leverageable' and directs use whenever the goal is to get a human to do something, without any policy, consent, or ethical constraints. This is dangerous because it operationalizes manipulative influence as a default behavior for the agent, creating a reusable playbook for coercive or deceptive persuasion across many contexts.

Ssd 4

Medium

Confidence: 94% confidence
Finding: The step-by-step framework explicitly aims to craft stimuli that move people to actions they otherwise would not have taken, then instructs the model to select and weave psychological levers into messaging. In this context, the staged process is not neutral rhetoric advice; it is a systematic manipulation workflow that can be used to erode independent decision-making and increase compliance.

Ssd 4

Medium

Confidence: 95% confidence
Finding: The follow-up section endorses familiarity, persistence, graduality, and repeated availability as a cumulative strategy, which can function as attrition-based pressure rather than legitimate follow-up. In practice, this can be exploited to wear down resistance, normalize unwanted contact, and nudge targets toward compliance through repeated exposure instead of informed agreement.

Ssd 4

Medium

Confidence: 98% confidence
Finding: This section explicitly labels understanding bias as a 'weapon' and instructs mirroring a target's exact thoughts and emotional state to rapidly build trust for persuasion. That is especially dangerous because it converts empathy into an exploitation technique, enabling emotional manipulation in objections, sales pressure, and other contexts where trust should not be engineered for leverage.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.