Deep Research Pro Litiao

• Outcome: Pass
• Latest audit: May 28, 2026, 10:53 AM
• Version:
  1.0.0

Overview

This is a web research helper whose search, page fetching, and report saving fit its stated purpose, though its documentation is inconsistent about Tavily/API-key use and local persistence.

Install only if you are comfortable with research queries going to external search providers or websites and reports being saved locally. Decide whether to provide TAVILY_API_KEY; without it, expect the DuckDuckGo fallback path. Avoid confidential, internal, or regulated topics unless those data flows and local files are acceptable.

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 (4)

Description-Behavior Mismatch

Medium

• Confidence: 94% confidence
• Finding: The README materially misrepresents operational requirements by claiming 'No API keys required' and 'Completely free' while the skill metadata indicates a preferred Tavily API path with DuckDuckGo only as fallback. This can mislead users about cost, data handling, and runtime behavior, causing them to install or invoke the skill under false assumptions.

Vague Triggers

Medium

• Confidence: 78% confidence
• Finding: The invocation examples are broad natural-language prompts such as 'Just ask your agent to research something,' which may cause the skill to trigger during ordinary conversation rather than only on explicit user intent. In an agentic system with web access and file-writing capability, ambiguous activation increases the risk of unintended external requests, data collection, or side effects.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The README advertises full-page fetching and saving results to files but does not warn users that queries and fetched URLs may disclose sensitive interests to third parties or that saved output may persist sensitive content locally. In a research skill, these behaviors are expected, but the absence of privacy and data-impact disclosure makes accidental data exposure more likely.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The skill directs the agent to save research outputs to a persistent local path without warning that user prompts, intermediate findings, or sensitive research topics may be written to disk. This can create unintended data retention, privacy issues, and later exposure to other local users, processes, backups, or indexing tools.

VirusTotal

VirusTotal findings are pending for this skill version.