Tp2
High
- Category
- MCP Tool Poisoning
- Confidence
- 85% confidence
- Finding
Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.
- Content
Security audit
Enterprise Bid Document AI — 企业招投标文书AI助手
Security checks for vulnerabilities and agentic risk
Overview
This skill is a disclosed bid-document assistant that handles sensitive business documents with explicit user warnings and human-review boundaries.
Before installing, confirm your organization allows third-party AI processing of bid materials, redact confidential pricing/client/personnel data, and treat generated bid/legal/business text as draft work requiring qualified human review. Use the pinned reviewed version and be cautious with any future added scripts or templates not present in this artifact.
Vulnerability Patterns
- Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
- Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
- Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
- Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
- Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
- MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)
Intent-Code Divergence
Medium
- Category
- Not specified by scanner
- Confidence
- 86% confidence
- Finding
The skill states that generated artifacts must be previewed and explicitly confirmed before any file write, yet it also enables
Writeand advertises generation workflows without consistently restating that guardrail. In a skill that processes bid documents containing confidential business data, this ambiguity can lead an agent to save sensitive or unsanitized content to disk prematurely, creating confidentiality and compliance risk.- Content
Autonomous Decision Making
Medium
- Category
- Excessive Agency
- Confidence
- 80% confidence
- Finding
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- Content
md - **Bid documents are confidential.** Uploaded files may contain pricing, technical proposals, and business-sensitive content. - **Sanitize before upload.** Remove or redact sensitive data (specific prices, client names, core technical parameters, personnel identity information) before submitting documents for analysis. - **No guaranteed deletion.** Uploaded documents are not guaranteed to be automatically deleted or anonymized after processing. - **Your responsibility.** You are responsible for ensuring your organization's data security policies allow using third-party AI tools for business document analysis. ### Quick Start
Autonomous Decision Making
Medium
- Category
- Excessive Agency
- Confidence
- 80% confidence
- Finding
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- Content
md - **Bid documents are confidential.** Uploaded files may contain pricing, technical proposals, and business-sensitive content. - **Sanitize before upload.** Remove or redact sensitive data (specific prices, client names, core technical parameters, personnel identity information) before submitting documents for analysis. - **No guaranteed deletion.** Uploaded documents are not guaranteed to be automatically deleted or anonymized after processing. - **Your responsibility.** You are responsible for ensuring your organization's data security policies allow using third-party AI tools for business document analysis. ### Quick Start
Intent-Code Divergence
Low
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
L694 acknowledges the category weights sum to 115% and instructs users to normalize by dividing by 1.15 or otherwise renormalize, while L696 says the total score is still calculated by the weighted sum with unchanged thresholds. These statements cannot both be true for the same model behavior, so the documentation contradicts itself about how the tool's risk assessment should work.
- Content