Back to skill

Security audit

Stock Investment System

Security checks for vulnerabilities and agentic risk

Overview

This is a Chinese stock-analysis framework with notable financial-advice risks, but I found no code execution, persistence, credential access, exfiltration, or hidden malicious behavior.

Install only if you want a Chinese-language stock-analysis checklist. Do not let it place trades or rely on it as licensed financial advice; verify all market data independently, treat 奇门遁甲 timing as non-evidence-based, and review any Wind/MCP integration before granting tool access.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Note
Location
references/wuwei-bayesian-v3.4.md:95
Finding

Invisible zero-width Unicode characters reduce instruction auditability

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
references/six-layer-matrix-v1.0.md:125
Finding

Standalone invisible Unicode character at end of reference file

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (14)

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/six-layer-matrix-v1.0.md (reported line 125)May include surrounding context.

md
#### 先验来源差异

| **​**  | **股票**             | **股权**            |
| ------ | ------------------ | ----------------- |
| P(好)来源 | 行业β + 市场情绪 + 分析师共识 | 行业 + 团队 + 退出 + 竞争 |
| 信息基础   | 公开信息·相对对称          | 尽调信息·高度不对称        |

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/wuwei-bayesian-v3.4.md (reported line 95)May include surrounding context.

md
#### 先验来源差异

| **​**  | **股票**             | **股权**            |
| ------ | ------------------ | ----------------- |
| P(好)来源 | 行业β + 市场情绪 + 分析师共识 | 行业 + 团队 + 退出 + 竞争 |
| 信息基础   | 公开信息·相对对称          | 尽调信息·高度不对称        |

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/wuwei-bayesian-v3.4.md (reported line 219)May include surrounding context.

md
#### 先验来源差异

| **​**  | **股票**             | **股权**            |
| ------ | ------------------ | ----------------- |
| P(好)来源 | 行业β + 市场情绪 + 分析师共识 | 行业 + 团队 + 退出 + 竞争 |
| 信息基础   | 公开信息·相对对称          | 尽调信息·高度不对称        |

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/wuwei-bayesian-v3.4.md (reported line 231)May include surrounding context.

md
#### 先验来源差异

| **​**  | **股票**             | **股权**            |
| ------ | ------------------ | ----------------- |
| P(好)来源 | 行业β + 市场情绪 + 分析师共识 | 行业 + 团队 + 退出 + 竞争 |
| 信息基础   | 公开信息·相对对称          | 尽调信息·高度不对称        |

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · references/wuwei-bayesian-v3.4.md (reported line 237)May include surrounding context.

md
#### 先验来源差异

| **​**  | **股票**             | **股权**            |
| ------ | ------------------ | ----------------- |
| P(好)来源 | 行业β + 市场情绪 + 分析师共识 | 行业 + 团队 + 退出 + 竞争 |
| 信息基础   | 公开信息·相对对称          | 尽调信息·高度不对称        |

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill provides actionable trading guidance such as opening, adding, reducing, and liquidating positions, including position sizing and execution gates, without an explicit warning that use may result in real financial loss. This increases the chance that users treat the output as authoritative operational advice rather than an educational framework, creating foreseeable harm in a high-stakes domain.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill goes beyond an investment-analysis framework and explicitly orchestrates another skill/tool (wind-mcp-skill) for data verification. Cross-skill delegation expands the trust boundary and can trigger unintended tool use, capability escalation, or hidden dependencies that are not declared in the manifest, making the skill harder to audit and sandbox safely.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill references a local file path (~/.hermes/references/skill-output-specs.md) as formatting guidance, creating an undeclared dependency on host-local state outside the skill package. In an agent setting, this can cause non-deterministic behavior, leakage of environment assumptions, or prompt/toolchain confusion if the file is missing, substituted, or contains attacker-controlled content.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document gives concrete trading decision instructions such as 建仓/加仓/减仓/清仓 and action thresholds like '至少四层🟢才动手' without any warning that the framework is not financial advice and may lead to losses. In the context of an investment skill, users could treat this as authoritative guidance and make real financial decisions without understanding the risk, suitability, or uncertainty involved.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The file explicitly recommends using 奇门遁甲 timing signals for buy/add/reduce/liquidate decisions and even provides operational rules by decision type, presenting a speculative divination method as part of a trading framework. Without a warning that such signals are unverified and unreliable for financial decision-making, users may be misled into taking materially risky actions based on pseudoscientific inputs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

该规则适用于所有文件类型。文档标题、说明、模板和操作指引均以中文固定呈现,未看到允许用户选择其他语言、或说明该技能仅面向特定中文使用场景的声明,构成潜在的语言/locale policy violation。

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The top-level activation cues are broad and include generic concepts like financial reports, volatility, announcements, catalysts, and abnormal information flow. In an agent skill, such loose triggers can cause unintended invocation during ordinary market discussion, leading the system to produce investment-analysis behavior outside the user's actual intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The event-driven checklist is triggered by vague conditions such as earnings, ±10% price movement, major announcements, industry catalysts, or abnormal information flow, without specifying scope, source, or disambiguation. This increases the chance that routine news or ambiguous chatter activates decision logic and produces actionable-looking investment recommendations unintentionally.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

SQP-3 applies to all file types for natural-language policy issues. The skill is entirely presented in Chinese and is targeted to A股/港股 investors, but it does not explicitly offer users a language preference or make clear that the locale-specific presentation is an intentional opt-in constraint.

Content

No source excerpt is available for this finding.