三國志略 / Histrategy

• Outcome: Pass
• Latest audit: Jun 18, 2026 at 9:01 PM UTC
• Version:
  2.0.0

Overview

This is a disclosed historical strategy game skill whose file storage, LLM use, and chat-based gameplay are aligned with its purpose.

Install if you want a persistent AI strategy game and are comfortable with local game saves and optional LLM API usage. Prefer explicit commands like /histrategy to avoid accidental activation, and delete saved rooms if you do not want gameplay history retained.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The trigger phrase "rome" is very broad and likely to appear in unrelated conversations, which can cause accidental invocation of the skill. In an agent environment, unintended activation can lead to unprompted file writes, API usage, or stateful game actions being performed in the wrong context.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The trigger "triumvirate" is ambiguous and may match normal historical, political, or educational discussion unrelated to the game. This increases the chance of accidental skill execution and any side effects associated with starting or resuming a stateful session.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill documents persistent on-disk storage of game state under the user's home directory but does not clearly warn about local file creation, retention, or possible sensitive content in logs. In multi-user or managed environments, silent persistence can expose gameplay history, prompts, and metadata beyond the user's expectations.

VirusTotal

64/64 vendors flagged this skill as clean.