ClawHub
Back to skill

Security audit

Multi-Source Research - 多源研究

Security checks across malware telemetry and agentic risk

Overview

This is a markdown-only research helper that matches its stated multi-source research purpose, with scope and privacy caveats but no hidden code or privileged behavior.

Install this for public or low-sensitivity research where broad external searching is intended. Avoid confidential, personal, or proprietary topics unless you are comfortable with queries being sent to search engines, academic sites, news sources, and social platforms; consider asking the agent to confirm before social-media monitoring or broad multi-source collection.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (3)

Vague Triggers

Medium
Confidence
88% confidence
Finding
The example trigger phrases are extremely generic everyday requests such as researching AI trends or finding papers, which can cause the skill to activate in situations broader than the author likely intended. In an agent ecosystem, overly broad activation language can lead to accidental invocation, unexpected access to external sources, and user confusion about which capability is handling the request.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The trigger condition is very broad: any request involving research, information gathering, literature review, or public-opinion analysis could activate this skill. Overbroad routing can cause the agent to invoke external-search behavior unexpectedly, increasing the chance of unnecessary data access, privacy leakage in queries, or use of untrusted sources when a narrower/local response would suffice.

Natural-Language Policy Violations

Medium
Confidence
82% confidence
Finding
The skill metadata and operating instructions are Chinese-only and do not offer language negotiation with the user. This can cause incorrect routing, misunderstood outputs, or hidden assumptions about source selection and jurisdiction, especially in multilingual deployments, though it is primarily a usability and safety-boundary issue rather than a direct exploit primitive.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.