Biaoshu Writer

• Outcome: Review
• Latest audit: May 28, 2026, 8:13 AM
• Version:
  5.4.0

Overview

The skill mostly performs bid-document generation, but it includes required AI-authorship masking and document metadata rewriting that users should review before installing.

Review before installing. Use only with bid files you are authorized to process, change output paths, use trusted package and font sources, and avoid the AI-trace removal or metadata-masking behavior for real submissions unless the tender rules and your organization explicitly allow it.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill clearly describes reading bid files and writing generated Word documents, but it does not declare permissions or prominently communicate those capabilities as formal access requirements. This can cause users or hosting systems to underestimate the skill's filesystem access, increasing the risk of unintended data exposure or unauthorized local file modification.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The installation/use instructions tell users to send tender documents to the AI, but provide no warning that bid files may contain confidential business information, pricing, personnel data, or other sensitive procurement materials. In the context of a tender-writing skill, this omission is more dangerous because the primary workflow explicitly involves uploading potentially sensitive source documents to an AI system, increasing the risk of unintended disclosure or improper handling.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill specifies a fixed output path under /Users/owen/Desktop/{项目名称}/ without warning the user or indicating that local files will be written there. Writing to a hard-coded desktop path can overwrite existing content, leak sensitive procurement data into an unexpected location, or fail unpredictably on other systems.

Ssd 2

Medium

• Confidence: 97% confidence
• Finding: The skill is explicitly framed as removing signs that text was AI-generated, which creates an evasion capability against AI-origin heuristics, review workflows, or institutional disclosure expectations. In context, this is more dangerous because the file provides systematic paraphrasing guidance aimed at preserving meaning while hiding provenance-related signals.

Ssd 2

Medium

• Confidence: 98% confidence
• Finding: These core instructions direct semantic-preserving rewrites specifically to strip identifiable AI patterns while maintaining content, which is effectively an evasion workflow. That increases misuse potential for deceptive submission of generated text in settings where provenance, authenticity, or disclosure matters.

Ssd 2

Medium

• Confidence: 97% confidence
• Finding: The repeated advice to inject first-person voice, emotion, uncertainty, and 'messiness' so writing sounds like a 'real person' goes beyond normal editing and supports impersonation of human authorship. In this skill context, that makes the capability more dangerous because it operationalizes deceptive stylistic masking rather than merely improving prose.

Ssd 4

Medium

• Confidence: 96% confidence
• Finding: The document provides a comprehensive, step-by-step taxonomy for removing AI markers across content, grammar, style, and conversational traces until output passes as human-written. This systematic guidance materially lowers the effort needed to evade AI-origin review and can facilitate plagiarism, academic misconduct, fraud, or policy circumvention at scale.

VirusTotal

VirusTotal findings are pending for this skill version.