Banshee's Last Cry Game CN

• Outcome: Pass
• Latest audit: May 27, 2026, 9:23 PM
• Version:
  1.0.0

Overview

This is a Chinese-language interactive horror text game that stores local gameplay progress, with no evidence of exfiltration, code execution, privileged access, or hidden install behavior.

Install only if you want a Chinese-language horror/mystery text adventure. It may create or update a local game_state.json save file with your gameplay choices and progress, so avoid entering real personal information as player notes or names.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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 (4)

Intent-Code Divergence

Medium

• Confidence: 88% confidence
• Finding: The skill claims to be a pure read-only status viewer, but the instructions also state that progress will be automatically saved to game_state.json. This creates a capability mismatch: callers may invoke the skill expecting no side effects, while the implementation behavior implies file modification. In an agent environment, misleading read-only claims can bypass user expectations, policy gating, or auditing around state-changing actions.

Natural-Language Policy Violations

Medium

• Confidence: 95% confidence
• Finding: The skill content is written entirely in Chinese and explicitly instructs the agent to operate in that language and narrative mode without checking the user's preferred language. This can override user expectations, reduce transparency, and create accessibility and safety issues if users cannot understand instructions, warnings, or game content well enough to give informed consent.

Natural-Language Policy Violations

Medium

• Confidence: 95% confidence
• Finding: The skill content is entirely written to drive Chinese-language interaction and does not offer any mechanism to detect or accommodate the user's preferred language. This can exclude or confuse users, reduce informed consent around the interaction mode, and create a poor or misleading user experience, though it is not a classic security exploit.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The invocation examples are broad natural-language phrases without explicit activation boundaries, which can cause the skill to trigger on ordinary conversation about characters rather than only on deliberate tool use. In an agentic system, this increases prompt-routing ambiguity and can leak gated narrative information or override other skills' intended handling of user requests.

VirusTotal

66/66 vendors flagged this skill as clean.