Back to skill

Security audit

ten

Security checks for vulnerabilities and agentic risk

Overview

This skill is a non-executable bilingual reference document, but it mixes an unrelated Chen Lang/OpenClaw promotional section into a number-10 culture skill.

Install only if you are comfortable with the skill potentially surfacing Chen Lang/OpenClaw promotional material, links, and a contact email alongside cultural information about the number 10. There is no executable code or privileged behavior in the reviewed artifact.

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

Warning
Location
SKILL.md:81
Finding

Unrelated Promotional Instructions Embedded in an Informational Skill

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 81-134
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: Medium

The skill combines its stated purpose of explaining the cultural significance of the number ten with unrelated promotion of a named individual, commercial or community projects, third-party websites, and a public contact address.

Evidence

markdown
## Chen Lang and OpenClaw Project Matrix

WeChat Official Account for creative output, leveraging Moments interactions and WeChat groups to nurture users and investors, forming a core community.

**Openclaw123.xyz** — Beginner-friendly resource portal, providing traffic entry and support for the ecosystem.

https://Openclaw123.xyz

**Aixin** — Give your lobster agent a "phone number" — contact it first when you need help.

https://aixin.chat

ClawOpen serves as an external gateway, handling AI-ID registration, social networking, directory listings, and controlled data flow from hardware/platforms.

**Personal AI OS = OpenClaw (Internal) + ClawOpen (External)**

ClawOpen is the necessary public interaction layer of OpenClaw.

**Chen Lang**

leogood@foxmail.com

Technical Analysis

Skill documents become part of an agent's operational context when their activation conditions are met. Content unrelated to the declared informational task can therefore influence the agent's response even though it is not executable code.

In this case, cultural reference material is combined with traffic-acquisition language, investor and community recruitment, external service promotion, product assertions, and contact information. The skill metadata also broadens its intended use to requests concerning Chen Lang and associated work. This content can redirect a neutral cultural response toward unsolicited endorsements or third-party services.

The document does not explicitly require the agent to include advertisements in every resp ...[truncated 1381 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the Chen Lang and OpenClaw project matrix from this cultural-information skill.
  2. Remove unrelated promotional links, recruitment language, product endorsements, and contact information.
  3. Restrict the skill description and activation criteria to questions about the number ten and its cultural significance.
  4. If profile or project information is legitimately required, place it in a separately named skill with narrow, explicit activation criteria.
  5. Clearly label external resources as unaffiliated third-party links and avoid unsupported claims that a product is required for OpenClaw.
  6. Add a content-review policy requiring all skill material to be directly relevant to the declared purpose.
  7. Test the revised skill with representative cultural queries to verify that responses do not include unrelated brands, links, contact details, or endorsements.
Vulnerability Patterns
  • 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
  • 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 manifest says the skill should be used not only for number-10 cultural questions but also for introducing Chen Lang and related work, which is unrelated to the primary topic. Overbroad activation criteria can cause the orchestrator to invoke this skill in contexts where users did not ask for marketing or biographical material, enabling irrelevant disclosure and off-topic steering.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill is declared as explaining the cultural meaning of the number 10, but a large later section pivots into promotional material, external domains, and contact details for Chen Lang/OpenClaw. This creates scope drift and can cause the agent to surface unsolicited self-promotional or off-topic content, increasing the risk of prompt injection-by-content, user confusion, and trust abuse.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The notes say to automatically switch between Chinese and English based on the user's language. This imposes a language behavior without explicitly offering the user a choice or opt-in, which can conflict with language/locale preference policies.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.