新闻联播政策摘要

• Outcome: Pass
• Latest audit: Aug 2, 2026 · 9:56 AM UTC
• Version: 1.0.0

Overview

This skill fetches public News Network Broadcast text and runs a local parser to summarize policy items; its behavior is mostly disclosed and purpose-aligned.

Before installing, be aware that this skill may run a bundled Python parser and create then delete a temporary transcript file. Use it for 新闻联播 policy summaries, and consider narrowing the broad "国家政策" trigger if accidental activation would be confusing.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Behavioral AST: exec() Call, eval() Call, Dynamic Import
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill directs the agent to write a local temporary file and run a local Python script, which are code and filesystem capabilities not declared in the skill metadata. This creates a transparency and control gap: reviewers and users cannot accurately understand what privileged actions the skill may take, increasing the chance of unintended local file access or execution.

Context-Inappropriate Capability

Medium

• Confidence: 92% confidence
• Finding: For a news-summary skill, instructing the agent to save fetched content to a local path and execute a local Python script expands the attack surface beyond what the user would reasonably expect. If the script or path handling is modified, compromised, or fed adversarial input, the agent could perform unintended local operations or execute unsafe code under the guise of simple content summarization.

Missing User Warnings

Low

• Confidence: 80% confidence
• Finding: The skill includes local file creation for fetched transcript content without informing the user, which weakens informed consent and obscures local data handling. Even if the content is not highly sensitive, hidden writes can create privacy, persistence, or forensic concerns on the host system.

Missing User Warnings

Low

• Confidence: 83% confidence
• Finding: The skill instructs subprocess execution of a local Python script without explicit warning, which conceals that code will be run on the host. Hidden execution is risky because users may expect a simple retrieval/summarization workflow, not local program execution with whatever privileges the agent possesses.

Missing User Warnings

Low

• Confidence: 78% confidence
• Finding: The skill deletes a local temporary file without notifying the user, which is a hidden filesystem side effect. Although cleanup is generally beneficial, undisclosed deletion can interfere with auditability, user expectations, or accidentally remove the wrong file if path handling is flawed.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The trigger phrase "国家政策" is broad and generic, so the skill may activate for many unrelated policy or news queries beyond the intended 新闻联播 use case. In a finance-oriented skill, overbroad invocation can cause unintended routing, user confusion, and accidental disclosure or generation of policy analysis where the user did not explicitly request this skill.

VirusTotal

60/60 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.