Evolver Plus

Security checks across malware telemetry and agentic risk

Overview

This skill is disclosed as a self-evolution tool, but it gives an agent broad automatic self-modification and rollback authority that users should review carefully.

Install only if you intentionally want an agent to propose and sometimes automatically apply persistent changes to its own memory and documentation. Before using it, narrow the allowed paths, require explicit approval for any overwrite or behavior-changing edit, disable unsupervised periodic review unless you truly need it, and verify backups before rollback.

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 (4)

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The skill is a self-modification engine and its activation criteria are defined with broad phrases like repeated errors, capability gaps, inefficiency, or receiving an evolution instruction. In this context, ambiguous triggers are dangerous because they can cause the agent to initiate self-alteration workflows without a clearly bounded scope, increasing the chance of unintended changes or abuse through crafted prompts.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The trigger list includes subjective conditions such as '效率明显低下', '能力缺失', and periodic self-review every 48 hours, but does not provide exclusion rules, confidence thresholds, or safeguards against adversarial prompting. Because this skill governs evolution proposals and possible file modifications, subjective triggers materially raise the risk of unnecessary or manipulated self-modification attempts.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The rollback trigger conditions are broad and operationally ambiguous, especially terms like '发现问题' and repeated failed evolutions, which can cause rollback logic to activate in situations that were not clearly intended. In a self-evolving agent context, unclear trigger criteria increase the risk of unnecessary or oscillating file restoration, potentially disrupting valid changes or masking root-cause issues.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The rollback snippet copies a backup directly over the target file without confirmation, integrity verification, path validation, or user-visible warning. In an automated self-modifying system, this can silently overwrite current state with an unintended or attacker-influenced backup, causing data loss, rollback abuse, or restoration of unsafe content.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal