NewsToday

Security checks across malware telemetry and agentic risk

Overview

NewsToday is a coherent news-briefing skill with expected web/news fetching and optional scheduled push behavior, though users should understand its broad triggers and third-party delivery channels.

Install if you want an agent-assisted news briefing tool that searches/fetches current news and can optionally schedule push notifications. Use push mode only with channels and user IDs you control, avoid sensitive tracked topics if third-party messaging retention matters, and disable scheduled notifications with the provided off command when no longer needed.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The header comment says there is no network activity, but the script emits instructions directing an agent to WebFetch multiple external URLs. In an agent-skill context, generated prompts are effectively operational behavior, so this mismatch can mislead reviewers and downstream policy checks about the skill’s real external-data access and trust boundary.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The trigger list is extremely broad and includes common phrases like '最新消息', '头条', 'trending', and 'breaking news', which can overlap with ordinary user requests and cause unintended skill activation. In an agent environment, this can misroute general queries into this skill, leading to unnecessary web access, unsolicited push-related flows, or incorrect handling instead of normal assistant behavior.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The '何时使用' section defines activation conditions using very broad examples such as '今天新闻', '有什么大事', and category-only requests like '科技' or '国际', which are hard to distinguish from normal conversation. This increases the chance the skill intercepts generic Q&A requests and performs aggregation/search behavior when the user did not clearly intend to invoke a dedicated news skill.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The keyword list is very broad and includes generic terms such as news summary, latest news, trending, AI news, and military news, which can match ordinary user conversation far beyond an intentional invocation of this skill. In an agent environment, this can cause over-triggering, accidental activation, and unsolicited retrieval or delivery of external content, increasing the chance of privacy leakage, spammy behavior, or prompt-routing hijack through ambiguous user phrasing.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal