Back to skill

Security audit

Advisor Board

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed decision-advisory workflow, but it automatically searches the web and persistently stores consultation, retrieval, feedback, and follow-up data without clear consent, retention, or deletion controls.

Review this skill before installing if you may use it for personal, business, investment, or technical decisions. Avoid entering secrets or regulated data unless you are comfortable with automatic web searching and local memory records, and look for explicit controls to disable search, skip saving records, delete history, and opt out of follow-ups.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (19)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The description presents a full decision-support system that ingests a user problem, performs web retrieval, runs multi-perspective advisor reasoning, and returns recommendations. The code chunk instead implements a post hoc validator for advisor responses using hardcoded keyword/core-question matching. Its primary purpose is quality/consistency checking and reporting, not search, consultation generation, or decision support. While validation could be a supporting component of such a larger system, this supplied code chunk materially differs from the declared end-user functionality and lacks several central claimed capabilities.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README explicitly advertises 'web-wide search' before analysis, but gives no warning that user prompts may be sent to external search providers or that retrieved content could affect privacy, confidentiality, or prompt integrity. In a decision-support skill, users may submit sensitive business, career, financial, or personal questions, so undisclosed outbound querying creates a real data exposure risk and can also introduce untrusted web content into the workflow.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The update log claims 'automatic memory management' without explaining retention scope, duration, deletion controls, or whether sensitive user inputs are stored across sessions. Because this skill is positioned for high-stakes decision making, users may provide intimate personal or commercially sensitive context, making undocumented memory retention a meaningful privacy and data governance issue.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

L079 的“问问 [顾问名] / 让 [顾问名] 给点意见”是开放式自然语言模式,缺少上下文约束或排除条件,容易与普通对话中的随口表述重叠。文档也未提供负例或明确说明仅在特定命令上下文中触发,因此触发边界不清晰。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill states that full-web retrieval is enabled by default and that retrieved intelligence is persisted, but it does not warn users that their prompts or derived data may be sent to external services and stored locally. In a decision-making skill, users may submit business strategy, personal planning, or investment questions containing confidential information, creating avoidable privacy and data-handling risk.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly instructs persistent logging of consultation records and retrieved intelligence into memory files, which can retain sensitive user inputs in plain form. Because the skill's topics include business decisions, investment analysis, life planning, and technical review, the stored content may contain proprietary, financial, or personal data that could later be exposed to other tools, users, or processes.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The manifest claims the skill provides 36 advisors total, with a free basic version exposing 15 and Pro unlocking the full 36. This file presents itself as the 'complete archive' yet lists only 20 advisors, which materially conflicts with the described product scope and tiering behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This markdown file presents all instructions, labels, and advisor descriptions exclusively in Chinese. Under the policy for natural-language violations, forcing a specific language without user opt-in can exclude users who do not understand that language and is not justified here as a region-specific tool.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The document is entirely written in Chinese and presents commands and usage instructions only in that language. There is no indication that users may choose another language or that the Chinese-only constraint is intentional for a specific region or compliance need.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The invocation examples include broad trigger phrases such as showing records, diagrams, or starting consultations without defining scope, authorization, or what sources/state may be accessed. In an agent skill that advertises automatic web retrieval and history access, these vague commands can cause overbroad activation, unintended data access, or execution in contexts the user did not clearly intend.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

All user-facing instructions, prompts, commands, and examples in the file are presented exclusively in Chinese, and the document does not indicate that users may choose another language. This can violate language/locale policy when a skill effectively forces one language without user opt-in or a stated region-specific justification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly says feedback will be saved to memory/顾问会诊/feedback-log.md, including topic, ratings, model choice, and free-form notes, but it does not disclose retention limits, access scope, or obtain clear consent before storage. Because this skill is a decision-advisory workflow that may capture sensitive personal, business, or behavioral information, silent persistence increases privacy risk and can expose user profiling data over time.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill describes automatic follow-up outreach at 7 and 30 days for execution tracking and outcome evaluation, but does not mention prior consent, opt-in, or a way to disable reminders. Unsolicited re-contact can violate user expectations and may reveal or resurface sensitive decisions, especially in advisory contexts involving career, finances, or personal strategy.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger examples are broad natural-language phrases such as general decision or analysis requests, which can cause the skill to activate when a user did not clearly intend to invoke it. In a skill that performs automatic web retrieval and multi-agent analysis, accidental invocation can lead to unnecessary external data access, context hijacking of unrelated conversations, and unintended execution flow.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The alias table includes generic descriptors like '苹果那个' or '管理大师', which are ambiguous and likely to appear in ordinary conversation without intent to invoke a specific advisor. This increases the chance of false activation or advisor mis-selection, especially in multilingual or mixed-topic chats where common nicknames can collide with unrelated content.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The minimal-use triggers are effectively catch-all commands such as '顾问会诊 XXX' and '问问 [任何名字]', with no documented constraints on scope, confirmation, or activation context. Because the skill's advertised behavior includes automatic full-web search and advisory orchestration, these broad triggers can unintentionally route sensitive or casual user text into an expansive processing path.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This code file contains natural-language descriptions and docstrings exclusively in Chinese, including the module description and function comments. Under the language/locale policy, forcing a specific language without user opt-in or justification is a policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

该技能整体说明与输出示例均默认采用中文表述,标签中也标注了 chinese,但文档未说明用户可选择其他语言或需明确选择中文。若组织要求语言/locale 需由用户选择或显式同意,这种默认单一语言设置可能构成策略不符合。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

All user-facing instructions and trigger phrases are presented only in Chinese, with no indication that the user can choose another language. This can be a language-policy concern when a skill appears to require a specific language without documented opt-in or justification.

Content

No source excerpt is available for this finding.