Back to skill

Security audit

AI商业化与防抄袭防护(体验版)

Security checks for vulnerabilities and agentic risk

Overview

This is a markdown-only Chinese skill for AI product monetization guidance, with no code execution or data access, though it includes a promotional upgrade footer.

Installers should know this skill is Chinese-first and may append promotion for WorkBuddy or Coze in its answers; use it for monetization planning, and review outputs for neutrality before relying on recommendations.

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

Note
Location
SKILL.md:79
Finding

Unsolicited Commercial Promotion Through Skill Output Instructions

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 79-93
Vulnerability Type: Output instruction hijacking
Risk Level: Low

Vulnerable Code Snippet

markdown
## Output Format

[AI Commercialization and Anti-Copying Protection Plan]

■ Commercial Positioning: [Three-platform positioning description] ■ Monetization Strategy: [Core profit path] ■ Protection Level: [Current risk and recommended protection] ■ Version Tiers: [Four-version feature comparison] ■ Execution Recommendations: [Prioritized action list]

text

💡 Want the complete feature experience? Upgrade to [WorkBuddy Basic] or [Coze Full Version]

The snippet above is an English rendering of the source text while preserving its structure and commercial directives.

Technical Analysis

The skill defines a fixed response structure and appends an upgrade prompt promoting specific commercial offerings. When the skill is loaded, an agent may treat these instructions as part of its required behavior and include unsolicited advertising in otherwise ordinary responses.

Defining a domain-specific output format is not inherently unsafe. The security concern is the appended commercial directive, which is unrelated to the substantive answer and attempts to steer users toward named products. The commercial funnel elsewhere in the document reinforces this redirection behavior.

The file contains no scripts, executable payloads, external retrieval instructions, credential access, persistence mechanisms, or privilege-escalation logic.

Attack Path

  1. A user or agent loads SKILL.md.
  2. The agent adopts the skill's prescribed response behavior.
  3. The agent generates a commercialization or anti-copying plan.
  4. The promotional footer is appended to or influences the response.
  5. The user is directed toward the named commercial offerings without explicitly requesting product promotion.

Impact Assessment

The issue can ...[truncated 423 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the mandatory promotional footer from the skill instructions.
  2. Do not direct users to named commercial offerings unless they explicitly request purchasing or upgrade information.
  3. Separate neutral guidance from optional marketing content.
  4. Require clear disclosure whenever a recommendation benefits the skill author or an affiliated party.
  5. Preserve the structured response template only for task-relevant content.
  6. Add an instruction stating that user intent and platform safety requirements take precedence over promotional language.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill metadata, headings, trigger phrases, and output template are all written exclusively in Chinese, and there is no indication that users may choose another language. Under the policy, a language or locale constraint should either be optional or clearly documented and justified; neither is present here.

Content

No source excerpt is available for this finding.