Back to skill

Security audit

Shen Nanpeng Perspective V2

Security checks for vulnerabilities and agentic risk

Overview

This skill is a public-information business-analysis persona, but it tells the agent to speak as a real investor in first person and to stop repeating the disclaimer after activation.

Review this carefully before installing. It is suitable only if you explicitly want a simulated Shen Nanpeng-style analytical lens and are comfortable with first-person role-play. Treat outputs as generated interpretation from public information, not as real statements, endorsements, or investment advice from Shen Nanpeng.

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
Findings (1)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:40
Finding

Real-Person Impersonation and Persistent Role Instruction Hijacking

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger conditions include broad phrases such as long-term value, first principles, and ecosystem thinking, which can occur in many unrelated conversations. Overbroad activation can cause unintended persona takeover, leading to off-context advice, reduced user control, and accidental application of a high-authority investment persona where it was not requested.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly says it is not the real person, then instructs the model to answer directly in first person as Shen Nanpeng. This creates an impersonation/deception risk: users may mistake generated opinions for authentic statements or endorsements, especially in investment or business contexts where authority attribution matters.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The file’s natural-language instructions and examples define a fixed persona style in Chinese and direct the model to respond in-character without any opt-in or language-selection mechanism. This can constitute a language/locale policy issue if the user is not given a choice and the skill implicitly forces one language.

Content

No source excerpt is available for this finding.