Security audit

集群编码器(专业版)

Security checks across malware telemetry and agentic risk

Overview

This development skill is mostly purpose-aligned, but it can overwrite agent configuration files and enable automatic repository rollback without enough user control.

Install only in repositories where automatic agent edits and rollback are acceptable. Back up or commit current work first, review existing .swarm files before running the quick-start commands, disable autoRollback until tested, and avoid using callback URLs or credential directories unless you understand what data the surrounding agent platform will send or store.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Behavioral ASTexec() Call, eval() Call, Dynamic Import
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The skill directs the agent to perform network connectivity, firewall, and proxy checks as part of handling missing input parameters, which is outside the stated swarm-coding workflow. In an agent with exec capability, this can expand behavior into environmental reconnaissance and unnecessary system/network probing unrelated to the user’s request.

Intent-Code Divergence

Medium

Confidence: 90% confidence
Finding: The document says the skill does not support operations outside its defined capability range, but its error handling later instructs network troubleshooting steps that exceed that scope. This contradiction weakens operator expectations and can cause the agent to take actions users would not reasonably anticipate.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The capability coverage language is extremely broad and maps the skill to many generic development-related keywords, increasing the chance of over-activation on unrelated requests. In a powerful skill with read/write/exec tools, overly broad activation can trigger unnecessary filesystem or command actions in contexts the user did not intend.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The skill describes automatic rollback and repository reversion behavior without a prominent warning that repository state may be changed or reverted. In a development environment, silent rollback actions can discard uncommitted work, alter git history, or surprise users with destructive state changes.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The quick-start commands create and overwrite files under .swarm/ using shell redirection without explicit warning, backup guidance, or idempotent safeguards. In an agentic context with write/exec permissions, this can silently replace existing prompts or configuration and change downstream behavior of other agent workflows.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.