企业年金查询技能

• Outcome: Review
• Latest audit: May 29, 2026 at 9:26 PM UTC
• Version:
  3.2.1

Overview

This skill is mostly a coherent company pension search tool, but it needs review because its scripts can write report and log files using unsanitized user-supplied company names.

Review before installing. Use it only for public-source company pension research, avoid sensitive or confidential target names unless you trust the configured search provider, and do not run it with company names or batch entries containing slashes, '..', or other path-like characters. Treat generated reports as templates requiring manual verification, and clear saved reports/cache when the investigation should not persist locally.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Intent-Code Divergence

Medium

• Confidence: 98% confidence
• Finding: The script generates a report template that asserts specific investigative methods, external-source verification, confidence levels, and conclusions, but the script itself only writes static markdown and performs placeholder substitution. This can mislead users into trusting fabricated due-diligence outputs, creating integrity and decision-making risk, especially if the report is later shared as if it were evidence-backed.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: Overly broad trigger phrases such as general welfare or benefits queries can cause the skill to activate on unintended requests. In this skill, unintended activation is more sensitive because the documented behavior includes external searches and report generation, which can lead to unnecessary data collection, network activity, and local file writes without clear user intent.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The skill describes generating and auto-saving investigation reports to local paths, but does not prominently disclose this side effect before use. This is dangerous because users may unintentionally persist potentially sensitive company research, query history, or collected source data on disk, especially in shared or monitored environments.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: Several triggers such as broad welfare/investigation phrases can match ordinary user requests that are not specifically about pension lookup, causing the skill to activate unexpectedly. In an agent system, overbroad activation can route unrelated user data or tasks into this skill, leading to inappropriate data collection, misleading responses, or unintended external searches.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The script sends user-supplied company names and related profiling queries to external search services such as Tavily or SearXNG without any explicit consent, warning, or data-handling notice. In an agent/skill context, this can leak sensitive investigative targets, internal interest, or confidential organization names to third parties and local linked skills outside the user's awareness.

VirusTotal

64/64 vendors flagged this skill as clean.