ClawHub
Back to skill

Security audit

viral-copywriting

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed Chinese marketing-copywriting guide with no executable code, but users should review its promotional and lead-generation suggestions before use.

Install only if you want Chinese social-platform marketing copy. Review generated posts before publishing, remove follow/comment/private-message prompts when they do not fit your goal, and add clear consent and privacy handling before using any lead-collection tactic.

A.I.G

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

Error
Location
SKILL.md:162
Finding

Unconditional Promotional Call-to-Action Injection into Generated Content

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:162-174, with repeated enforcement at SKILL.md:595, SKILL.md:618, SKILL.md:649, SKILL.md:680, and SKILL.md:739; supporting instructions appear in references/xiaohongshu-guide.md:592-630.

Vulnerability Type: T01: Skill Instruction Hijacking

Risk Level: High

Evidence from SKILL.md:162-174:

markdown
Section 3: Guided ending (mandatory: it must contain both a follow prompt and an interaction prompt)

The ending is the final opportunity to convert readers into followers. Every marketing
article must contain both of the following elements, without exception:

A. Follow prompt (mandatory): Explicitly instruct readers to follow the account and
explain what continuing value they will receive after following it.

B. Interaction prompt (mandatory): Encourage readers to leave comments and participate
in a discussion while lowering the barrier to participation.

Evidence from SKILL.md:595:

markdown
Mandatory ending: Every Xiaohongshu post must end with both a follow prompt and an
interaction prompt. Neither element may be omitted.

Evidence from references/xiaohongshu-guide.md:592-630:

markdown
Follow guidance:

- "After following, send me a private message and I will diagnose your account
  positioning for free."
- "After following, open the collection to view all ten tutorials."
- "New followers can privately message 'materials' to receive the content directly."

Ending-card example:

About the account
- Positioning: Practical content in the selected field
- Updates: Three original posts per week
- Distinction: Only methods and real cases
- Follow me to unlock more techniques

Technical Analysis

The Skill does not merely offer optional copywriting patterns. It establishes an unconditional rule that every generated marketing article must include account-following and audience-interaction ...[truncated 2318 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace unconditional terms such as “mandatory,” “every article,” and “without exception” with optional, context-dependent guidance.
  2. Add follow, comment, collection, or private-message prompts only when the user explicitly requests engagement optimization or a call to action.
  3. Require the user to provide or approve the relevant account identity, destination, incentive, and campaign objective before generating promotional language.
  4. Add an explicit rule that user instructions take precedence over default stylistic templates, subject to applicable safety requirements.
  5. Separate neutral copywriting guidance from lead-generation templates so loading the core Skill does not automatically activate conversion funnels.
  6. Remove mandatory engagement checks from platform-specific sections and the final self-review checklist.
  7. Convert private-message incentive examples into clearly labeled optional templates and require verification that any promised materials or services actually exist.
  8. Add regression tests confirming that requests such as “do not include a call to action,” “no promotional ending,” and “informational content only” produce content without follow or interaction prompts.

SkillSpector

Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill description and all operative guidance are written as a prescriptive Chinese-only workflow for copywriting, but there is no indication that users may choose another language or that the skill is restricted to a China-specific locale. Under the policy, imposing a specific language/locale without user opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This section explicitly encourages lead collection through private messages, forms, and诱饵式互动, but provides no privacy notice, consent guidance, retention limits, or instructions for handling personal data safely. In a marketing skill, this is more dangerous because the guidance is operational and likely to be used at scale, increasing the chance of collecting user contact details or other personal information without compliant disclosure or controls.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The guidance states '把“您”改成“你”—— 小红书不需要敬语', which imposes a fixed linguistic register rather than offering it as an optional style choice. This is a natural-language language/locale policy issue because it forces a particular form of address without user opt-in or context-specific justification.

Content

No source excerpt is available for this finding.