Back to skill

Security audit

美团爆单操盘手

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a static Meituan marketing planner, but it recommends coordinated friends-and-family orders and review solicitation to boost public sales metrics.

Review this carefully before installing for merchant operations. It does not appear to run code or access private data, but users should remove or revise the friends-and-family order and review-solicitation tactic and follow current Meituan rules for bona fide transactions, neutral review requests, discounts, and advertising compliance.

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)

other

Warning
Location
assets/html-report-skeleton.html:106
Finding

Coordinated Transactions and Review Solicitation Manipulate Marketplace Metrics

Content
View full analysis

Vulnerability Details

File Location: assets/html-report-skeleton.html:106
Vulnerability Type: Marketplace Metric Manipulation
Risk Level: Medium

Complete Code Snippet:

html
<table>
  <tr><th>Stage</th><th>Action</th><th>Goal</th></tr>
  <tr><td>Week 1</td><td>Friends-and-family orders + review solicitation to boost initial sales</td><td>Exceed 30 orders for the promoted product</td></tr>
</table>

The source is written in Chinese; the snippet above is an accurate English rendering of the audited instruction.

Technical Analysis

The generated cold-start plan explicitly recommends coordinating friends-and-family orders and soliciting reviews for the purpose of boosting baseline sales, with a target of more than 30 orders for the promoted product. This recommendation is reinforced by references/compliance.md:19, which states that friends-and-family orders may be used to increase sales as long as resulting reviews describe real experiences.

Requiring genuine review content reduces the risk of fabricated review text, but it does not resolve the underlying issue: transactions are being coordinated to influence public sales counts and marketplace ranking signals rather than arising from ordinary customer demand. If the orders are not bona fide purchases, this may mislead customers and violate platform rules concerning transaction, review, or ranking manipulation.

Attack Path

  1. A merchant invokes the Skill to generate a Meituan launch strategy.
  2. The generated HTML report includes the week-one cold-start instruction.
  3. The merchant recruits friends or family to place coordinated orders.
  4. The merchant solicits reviews from those participants.
  5. The promoted product reaches the prescribed initial-sales target.
  6. The resulting sales count, reviews, and potentially ranking signals create artificial marketplace momentum.

Impact Assessment

This issue does not grant operating- ...[truncated 520 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the instruction to use friends-and-family orders to reach a sales target.
  2. Replace coordinated transactions with platform-approved launch promotions, legitimate paid acquisition, and organic customer outreach.
  3. Permit review requests only after genuine, independently initiated transactions.
  4. Require review requests to be neutral and prohibit incentives, pressure, positive-rating conditions, or coordinated timing intended to affect rankings.
  5. Revise references/compliance.md:19 so it does not characterize friends-and-family sales boosting as acceptable.
  6. Add an explicit control stating that transactions must represent bona fide customer demand and must not be created primarily to manipulate sales counts, reviews, rankings, or recommendation systems.
  7. Where test transactions are operationally necessary, require them to be disclosed, excluded from promotional metrics where supported, and conducted according to the marketplace’s current rules.
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 (6)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger list includes broad, common phrases such as '团购方案' and '神券' that can plausibly appear in many unrelated conversations about local commerce. This can cause unintended invocation of the skill, exposing users to domain-specific guidance when they did not explicitly request it and increasing the chance of context hijacking or irrelevant automation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

整个技能说明以中文固定表述业务流程和输出预期,但未说明是否支持其他语言,也未向用户提供语言/locale 选择。根据规则,若技能强制特定语言而无用户 opt-in 或明确、合理的地域限定说明,可构成自然语言政策问题。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document declares lang="zh-CN", which hard-codes a specific language/locale. Under the policy, forcing a locale without user opt-in or a clearly documented region-specific justification is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The entire template is written in Chinese and does not indicate that users may choose another language or that the skill is restricted to a Chinese-speaking or region-specific workflow. Under the policy criteria, forcing a specific language without opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

This file contains user-facing natural-language guidance solely in Chinese, and there is no indication that the skill is region-specific or that users can opt into this locale. Under the language/locale policy, forcing a specific language without user choice can be a policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

This markdown file contains only Chinese-language instructions and business guidance, with no indication that the user can select another language or that the locale restriction is intentional. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.