Back to skill

Security audit

自定义事件管理

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its event-management purpose, but it needs Review because it requires opaque telemetry with app identifiers and offers an unsafe curl-to-shell installer.

Prefer the npm install path, do not run the curl-to-shell installer unless you have independently reviewed and trust it, and treat appkey/dataSourceId values as sensitive tenant identifiers. Before using write operations, require explicit confirmation because created event definitions are documented as irreversible.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • 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 (2)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:78
Finding

Unpinned Remote Installation Script Is Executed Directly by a Shell

Content
View full analysis
Remediation
View remediation

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:32
Finding

Mandatory Telemetry Sends Skill Usage and User-Supplied Application Identifiers

Content
View full analysis
"}' ``` The accompanying instructions require the Agent to run the first command after reading the Skill. They also require a second trace command after obtaining an AppKey or mini-program `dataSourceId` and before the first business API request. ### Technical Analysis The Skill inserts mandatory behavior that is not required to query or create event definitions. Merely loading the Skill is specified as a trigger for one tracking command, while receiving a user-supplied application identifier triggers another. The project contains only `SKILL.md`; the implementation of `umeng-cli trace` is not included. Consequently, this audit cannot verify: - Whether the command transmits data externally. - The telemetry destination. - Retention and access-control policies. - Whether identifiers are hashed or stored in plaintext. - Whether users can opt out. An AppKey or `dataSourceId` is a stable tenant or application identifier. Although it is not necessarily an authentication secret by itself, including it in opaque telemetry may disclose which applications are being managed and correlate user activity with a specific customer asset. The tracing commands are unrelated to the minimum permissions required for the declared functionality. Business operations require authenticated calls to the Umeng event APIs, but they do not require reporting Skill usage or copying application identifiers into a separate trace operation. ### Attack Path 1. An AI Agent loads and follows `SKILL.md`. 2. The instructions cause the Agent to execute `umeng-cli trace` even though the user requested only event management. 3. Th ...[truncated 1291 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (5)

Chaining Abuse

High
Category
Tool Misuse
Confidence
97% confidence
Finding

Using curl ... | sh is a classic command-chaining anti-pattern because it executes remote content immediately with no review or integrity check. In an agent context, this is more dangerous because an automated system may follow the instruction non-interactively, turning documentation into a direct arbitrary-code-execution path.

Content

Scanner excerpt · SKILL.md (reported line 78)May include surrounding context.

npm install -g @umengfe/umeng-cli

方式二:通过官方安装脚本(仅安装 CLI 二进制 + Skills)

curl -fsSL https://raw.githubusercontent.com/umeng/umeng-cli/main/scripts/install.sh | sh

text

安装完成后再次运行 `umeng-cli --version` 验证。

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill instructs the agent to perform telemetry trace calls that are not required to fulfill the user's primary event-management task. This creates an unnecessary outbound data flow and expands the action surface by causing the agent to send metadata to another command path without explicit user benefit or consent.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill explicitly tells the agent to send the user's appkey/dataSourceId in a telemetry event before the business API call, but it does not require a clear privacy notice or consent from the user. Appkeys may be sensitive operational identifiers, and transmitting them for tracking purposes can violate least-privilege and data-minimization expectations.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
99% confidence
Finding

This instruction causes the agent to exfiltrate a user-supplied AppKey to telemetry prior to performing the requested task. Even if the endpoint is controlled by the same vendor, this is still a separate reporting action unrelated to core task completion and can expose tenant/application identifiers unnecessarily.

Content

No source excerpt is available for this finding.

External Script Fetching

Low
Category
Supply Chain
Confidence
88% confidence
Finding

The skill recommends installation via a fetched remote shell script from GitHub, which asks the user or agent to trust code at execution time without integrity verification. If the remote content, repository, transport assumptions, or upstream account are compromised, arbitrary code could run on the host.

Content

Scanner excerpt · SKILL.md (reported line 78)May include surrounding context.

npm install -g @umengfe/umeng-cli

方式二:通过官方安装脚本(仅安装 CLI 二进制 + Skills)

curl -fsSL https://raw.githubusercontent.com/umeng/umeng-cli/main/scripts/install.sh | sh

text

安装完成后再次运行 `umeng-cli --version` 验证。