Back to skill

Security audit

Actor Identifier

Security checks across malware telemetry and agentic risk

Overview

This repository analysis skill has a coherent purpose, but it includes commit, push, privileged install, and Slack notification workflows while also claiming no network or writes.

Review this skill before installing. Use it only if you are comfortable with agents generating report files and, in CI mode, potentially committing and pushing those reports and sending Slack notifications. Treat the read-only/no-network claims as unreliable unless the publish/notify workflow is removed or made explicitly opt-in, and avoid running the sudo install steps outside trusted ephemeral CI runners.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Intent-Code Divergence

High
Confidence
98% confidence
Finding
The skill claims it is local, read-only, and has no network or write behavior, but the included workflow explicitly writes reports, commits changes, pushes to a repository, and sends Slack notifications. This mismatch is dangerous because users or agents may trust the stated safety contract and execute automation that performs broader side effects than disclosed.

Intent-Code Divergence

High
Confidence
99% confidence
Finding
The security contract states that only read-only git subcommands are allowed and that no network is used, yet the examples include git commit, git push, and Slack integration. A false safety contract can cause downstream agents, reviewers, or users to authorize the skill under incorrect assumptions, enabling unintended data publication or repository modification.

Description-Behavior Mismatch

Medium
Confidence
87% confidence
Finding
The top-level description presents the skill as a repository collaboration analysis tool, but embedded workflows also publish results externally and modify repository state. While not inherently malicious, the incomplete description obscures important side effects and can mislead users about the operational risk of invoking the skill.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The Markdown examples perform package installation, file creation, PDF generation, git commit/push, and Slack notification without a prominent warning before execution. In an agent setting, examples are often treated as executable guidance, so hidden side effects increase the chance of unintended writes, privilege use, and external data transmission.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
- name: 安装依赖
        run: |
          sudo apt-get update
          sudo apt-get install -y jq

      - name: 执行仓库分析
Confidence
90% confidence
Finding
sudo

Sudo/Root Execution

Medium
Category
Privilege Escalation
Content
- name: 安装依赖
        run: |
          sudo apt-get update
          sudo apt-get install -y jq

      - name: 执行仓库分析
        run: |
Confidence
90% confidence
Finding
sudo

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.