Back to skill

Security audit

Sheng Xitai Perspective V2

Security checks for vulnerabilities and agentic risk

Overview

This is a single Markdown persona-style investment perspective skill with no code execution, persistence, or data access, but users should treat it as a simulated viewpoint rather than real affiliation or financial advice.

Install this only if you want the assistant to answer some investment and industrial-strategy questions through this named investor-style lens. Treat outputs as a simulated framework for thinking, not professional financial advice or confirmation that the assistant represents the named person or Hongtai Fund.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (3)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger list includes broad, high-frequency domain phrases such as '早期投资', '产业生态', and '硬科技投资', which can cause the skill to activate in ordinary finance or strategy discussions without the user explicitly requesting this persona. That creates unintended instruction takeover risk, where the assistant may switch into a biased role and constrain responses based on the skill's framing rather than the user's actual intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The scenario-based activation rules are loosely defined and lack boundaries or exclusions, so common user questions like whether an early-stage project is worth investing in may automatically invoke the skill even when the user did not ask for that perspective. This increases the chance of non-consensual persona steering and hidden prompt interference in normal conversations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill prescribes a fixed communication style, including first-person claims like '我的经验' and '洪泰的实践', without user language or attribution controls. This can mislead users into believing the model has real-world identity, direct experience, or organizational affiliation, increasing deception and provenance risks.

Content

No source excerpt is available for this finding.