合同文档转写技能

• Outcome: Review
• Latest audit: May 29, 2026, 9:29 PM
• Version:
  1.0.0

Overview

This skill is a local contract-to-Word generator, but it should be reviewed because it can pick the wrong Desktop file and generate authoritative-looking output with hard-coded or incomplete contract details.

Install only after reviewing the behavior. Run it on a copy of the intended contract, confirm the exact input file and output path, and manually verify contract number, parties, requirements, and generated diagrams before relying on the Word document. The author should remove hard-coded contract data, require explicit file paths, disclose and pin dependencies, and make the implementation match the stated workflow.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• 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)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 90% confidence
• Finding: The skill advertises and orchestrates local file reads, local file writes, and shell-capable automation, but does not declare permissions or clearly constrain those capabilities. This is dangerous because an agent or user may authorize the skill without understanding it can access desktop files and invoke external tooling, increasing the chance of unintended data exposure or local system side effects.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 95% confidence
• Finding: The skill description promises contract-driven document conversion, architecture generation, and six prototype images, but the reported behavior instead relies largely on filename-derived values, template content, and external screenshot commands. This mismatch is security-relevant because users may provide sensitive contracts under the assumption of bounded processing, while the actual implementation uses undeclared tooling and produces misleading outputs that can hide data handling or execution risks.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The skill instructs reading a contract from the user's desktop and writing a generated document back to the desktop without an explicit warning or consent flow for local file access and modification. In context, contracts commonly contain sensitive business data, so silent access to desktop files increases the risk of privacy violations, accidental overwrites, or processing of the wrong document.

Missing User Warnings

Medium

• Confidence: 84% confidence
• Finding: The skill describes automated screenshot generation and execution of a script, including browser automation via Playwright, without warning the user that local script processing and external command execution will occur. This matters because automation tooling can fetch dependencies, launch browsers, consume system resources, and create files, all of which expand the attack surface beyond simple document transformation.

VirusTotal

66/66 vendors flagged this skill as clean.