paper-summary-scripted

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:41 PM UTC
• Version:
  1.0.0

Overview

The skill openly runs a local preprocessing script for user-selected papers and arXiv PDFs, with no evidence of hidden, destructive, or unrelated behavior.

Install only if you are comfortable running a bundled Python script on selected papers. Avoid passing private or unrelated local files, and remember that downloaded PDFs, extracted text, and combined text are stored under the output directory and may be used in the configured model workflow.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (3)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill directs the agent to execute a Python script that reads local files, writes outputs, and performs network downloads, yet the skill declares no permissions. This creates an authorization and transparency gap: a caller or platform may treat the skill as low-privilege while it can access local paper files and fetch remote content, increasing the risk of unintended data access or egress.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 84% confidence
• Finding: The documented behavior promises a four-stage summarization and verification workflow, but the detected implementation only preprocesses/downloads files and combines extracted text. This mismatch is security-relevant because users and orchestrators may trust the skill to perform bounded summarization while it actually introduces file/network side effects without delivering the claimed review stages, which can conceal unexpected data handling and break downstream safety assumptions.

Natural-Language Policy Violations

Medium

• Confidence: 92% confidence
• Finding: The template claims the output should use the user's requested language, but the fixed Chinese section headings create contradictory instructions that can cause the agent to ignore user language preferences or produce mixed-language output. This is a specification-integrity issue rather than code execution, but it can degrade reliability, confuse downstream consumers, and break workflows expecting consistent localization.

VirusTotal

67/67 vendors flagged this skill as clean.