Back to skill

Security audit

legal-lingo-translator

Security checks for vulnerabilities and agentic risk

Overview

This text-only legal translation skill has proportionate capabilities, but its examples can introduce unsupported facts that users should correct before use.

Use it as a drafting aid. Check every factual statement against your original account, especially claims about repeated demands or deliberate refusal to repay. Verify deadlines and obtain qualified legal review before submitting documents.

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
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

The forward-translation example adds unsupported factual assertions despite the explicit prohibition on fabricating evidence.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The loan example adds “经本人多次催告” (after repeated demands) and changes nonpayment into “拒不返还” (refusal to repay), although neither assertion is established by the source text. This contradicts the skill’s no-fabrication rule and may teach the agent to introduce unsupported facts into copy-ready legal documents, potentially undermining the user's credibility or misrepresenting events. The legal-document context increases the consequences of this output-integrity flaw, despite the otherwise explicit safeguards.

Content

No source excerpt is available for this finding.