麦赛尔夫

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed investment-research template that writes sourced research reports and workspace notes, with no evidence of hidden access, credential use, exfiltration, or destructive behavior.

Install this if you want an agent that creates Markdown investment research reports and local workspace logs from public financial sources. Review the strict English-path and report-language defaults before use, keep the workspace scoped, approve any package installation yourself, and treat outputs as research references rather than investment advice.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Natural-Language Policy Violations

Medium

Confidence: 90% confidence
Finding: The skill hard-codes a Chinese acknowledgment phrase for every task, regardless of the user's language or locale. This can cause instruction mismatch, reduce usability, and potentially confuse downstream workflows or users who expect responses in another language, though it does not directly enable code execution, data exfiltration, or privilege escalation.

Natural-Language Policy Violations

Medium

Confidence: 95% confidence
Finding: The skill imposes an English-only filename/path restriction as a hard rule without user consent or a documented technical necessity. This can override user preferences, exclude legitimate non-English workflows, and cause inappropriate handling of user data or outputs in multilingual environments.

Natural-Language Policy Violations

Low

Confidence: 88% confidence
Finding: Requiring the report body to be in the principal's language without explicit user choice can override the active user's preferences and reduce transparency or usability for other stakeholders. In a shared or delegated workflow, this may lead to misdelivery, misunderstanding, or outputs that do not match the requester's intent.

Natural-Language Policy Violations

Medium

Confidence: 87% confidence
Finding: The skill hard-codes an English-only filename/path requirement, which can override user intent, exclude non-English workflows, and create unnecessary restrictions unrelated to security. In an agent setting, this can lead to policy non-compliance, mishandling of user-specified filenames, or silent renaming that breaks expected file locations and downstream automation.

Natural-Language Policy Violations

Medium

Confidence: 92% confidence
Finding: The document explicitly mandates "English file paths only" because "CJK characters in filenames cause encoding issues across systems and tools." Even if operationally motivated, this is a restrictive policy that can exclude non-English users and locales without documenting a narrowly scoped technical requirement or safer fallback behavior, making it a genuine policy-level security/usability issue.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal