Back to skill

Security audit

Investment Cognition Detector

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only investment self-assessment skill with disclosed public web research and no code execution, persistence, credential access, or account-changing authority.

Install only if you want a Chinese-language workflow that uses public web research to challenge your understanding of a company. Do not treat its scoring as financial advice, and avoid sharing private portfolio details unless you are comfortable with them being used in the chat and potentially in web-search context.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

整个技能说明、触发语和输出要求均以中文固定表述,且未说明可根据用户偏好切换语言。这可能构成语言/locale 强制约束,而文档中也没有区域合规或受众限定的正当化说明。

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger condition includes a broad catch-all phrase for 'any request' related to testing investment understanding, which can overlap with normal financial conversation and cause unintended invocation. In a finance-oriented skill, accidental activation can lead the agent to start collecting detailed investment views and performing web research without the user explicitly opting into this workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The natural-language instructions and output expectations are entirely Chinese, and there is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-language audience. Per the policy, forcing a specific language without user opt-in is a language/locale violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill instructs the agent to use web_search and web_fetch to research a company but does not disclose to the user that external network access or third-party data will be used. This reduces transparency and can surprise users, especially when their query appears conversational but results in external lookups and potentially stale or inaccurate financial information influencing the output.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The workflow explicitly states the AI will research the company using public materials, but the skill does not warn the user that their query may trigger external web access. This can create a transparency and privacy issue because users may not expect outbound retrieval based on the company or context they provide, even if the sources are public.

Content

No source excerpt is available for this finding.