Security audit

Advisor Board

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent decision-support skill that openly uses web search and memory, but users should be careful with sensitive information.

Install only if you are comfortable with a skill that may perform web searches and save consultation history, retrieved intelligence, feedback, and follow-up notes in agent memory. Avoid entering secrets, confidential business plans, regulated personal data, or sensitive financial details unless you know how your OpenClaw environment handles web queries, retention, and deletion.

SkillSpector

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 (8)

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The README advertises '全网检索增强' and '自动记忆管理' but does not disclose what user data may be transmitted, stored, retained, or reused. In a decision-support skill, users may submit sensitive business, financial, or personal information, so undisclosed search and memory features create a real privacy and data-handling risk even if no explicit exploit code is present.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill states that web search is enabled by default and that consultation/search records are saved to persistent memory paths, but it does not clearly obtain consent or disclose what user data will be stored and retained. This can lead to unintended collection and retention of sensitive business, investment, personal, or strategic information entered by users.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger language uses broad scenario labels such as strategic direction, major decisions, competition strategy, and market research without clear boundaries, exclusions, or disambiguation rules. In an agentic system, this can cause over-broad activation and routing of the skill on unrelated or sensitive user requests, increasing the chance of inappropriate delegation, misleading advice, or prompt-scope interference.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: Multiple sections define activation through generic business contexts like business model, strategic planning, marketing strategy, and product launch, but do not specify what user intent or evidence is needed to invoke the skill. This ambiguity can let the system treat ordinary conversation as a match, causing accidental activation, context bleed, or overreach into domains the skill is not equipped to handle safely.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The quick-reference mapping routes everyday scenarios like strategy, execution, product design, and investment judgment directly to advisors with no specificity, confidence thresholds, or safety gates. If used by an orchestrator, this creates a broad attack surface for unintended routing and can amplify bad decisions by presenting generalized heuristics as authoritative guidance in ambiguous contexts.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The documented triggers are broad, conversational phrases like '我该不该做 XXX？' and '帮我全面分析一下' that overlap with ordinary user speech. In an agent environment, this can cause unintended invocation or over-invocation of the skill during normal conversation, potentially triggering web retrieval and multi-step analysis when the user did not explicitly intend to activate this skill.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The alias-based activation guidance allows vague descriptors such as '苹果那个' or '那个搞第一性原理的', which are highly ambiguous and may match unintended entities or conversational references. This weakens invocation integrity and increases the chance that the wrong advisor flow is selected without clear user consent.

Vague Triggers

Medium

Confidence: 97% confidence
Finding: The shortcut '问问 [任何名字]' is explicitly underspecified and effectively invites arbitrary free-text matching against names. In practice, this can cause accidental activation, incorrect routing, or manipulation through crafted input that resembles a name, especially in multi-agent or tool-using systems.

Static analysis

No suspicious patterns detected.