Back to skill

Security audit

中国机动车驾驶证识别与抽取(ADP)

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to perform the advertised driving-license extraction, but its installation guidance and bundled CLI documentation expose more authority than the narrow skill needs.

Review before installing. Prefer the npm package or a versioned release, avoid curl|bash or iex installers, use a least-privilege account, and only provide driving-license files and API credentials you are comfortable sending to Laiye ADP. Do not use the broader custom-app or generic document commands unless you intentionally want that wider ADP CLI capability.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • 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)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:64
Finding

Unpinned Remote Installer Scripts Are Downloaded and Executed Directly

Content
View full analysis

Vulnerability Details

File Locations:

  • SKILL.md:64
  • SKILL.md:69
  • README.md:30
  • README.md:35
  • README-CN.md:29
  • README-CN.md:34

Vulnerability Type: Unverified remote payload retrieval and execution
Risk Level: High

Complete Vulnerable Code Snippet:

bash
# Shell installation method for Linux and macOS
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
powershell
# PowerShell installation method for Windows
irm https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.ps1 | iex

Technical Analysis

The installation instructions retrieve scripts from the mutable main branch of an external GitHub repository and immediately execute the responses using bash or PowerShell Invoke-Expression.

The downloaded scripts are not included in the audited project. Consequently, their effective behavior cannot be determined from this project and may change after the Skill has been reviewed. The instructions do not:

  • Pin the script to an immutable commit or signed release.
  • Verify a cryptographic signature.
  • Validate a published checksum.
  • Save the script for inspection before execution.
  • Restrict execution through a sandbox or reduced-privilege account.

HTTPS protects the transport connection but does not protect against malicious or unauthorized changes made at the upstream repository. Compromise of the repository, maintainer account, branch, or release process could replace the installer with an arbitrary payload.

The upstream location is presented as the vendor's official CLI repository, and no evidence in the audited files proves that it currently serves a malicious script. Nevertheless, directly executing mutable remote content creates a remote code execution channel and is not necessary for the Skill's declared document-recognition functionality.

Attack Path

  1. An attacker compromise ...[truncated 1527 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove all pipe-to-shell and download-to-Invoke-Expression installation instructions.
  2. Publish installers as versioned release artifacts rather than retrieving them from a mutable branch.
  3. Pin each download to a reviewed, immutable version or commit.
  4. Publish SHA-256 or stronger checksums through an independently protected release channel and require users to verify them before execution.
  5. Cryptographically sign release artifacts and document signature verification using a pinned signing identity.
  6. Download the installer to a local file first, verify its checksum and signature, and only then execute it.
  7. Prefer a version-pinned package-manager command or signed prebuilt binary. Avoid global installation where a project-local installation is sufficient.
  8. Run installation and document processing under a dedicated, non-administrative account with access only to the required input and output directories.
  9. Update SKILL.md, README.md, and README-CN.md consistently so no insecure fallback remains.
  10. Document that driving-license files are transferred to an external cloud service and require explicit user authorization, appropriate data-retention controls, and compliance with applicable privacy and data-residency requirements.
Vulnerability Patterns
  • 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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (18)

Chaining Abuse

High
Category
Tool Misuse
Confidence
99% confidence
Finding

The shell pipe into bash is a classic chaining-abuse pattern because it transforms untrusted remote content directly into executable commands with no validation boundary. In this skill context, that is especially risky because the README is positioned as setup guidance, making copy-paste execution likely and enabling rapid compromise if the remote content is altered.

Content

Scanner excerpt · README-CN.md (reported line 29)May include surrounding context.

bash
# 方法 2: Shell 脚本(Linux / macOS,无 npm 环境时使用)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
bash

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The README presents a broad, general-purpose ADP document-processing CLI with parsing, extraction, batch processing, task querying, and multi-scenario support, which materially exceeds the declared skill scope of Chinese driving-license recognition and field extraction. In an agent setting, this scope mismatch is dangerous because it can induce the agent to invoke capabilities on arbitrary local files, remote URLs, or unrelated document types, expanding data access and processing beyond user expectations.

Content

No source excerpt is available for this finding.

Chaining Abuse

High
Category
Tool Misuse
Confidence
96% confidence
Finding

The | bash pattern is a classic command-chaining abuse vector because it executes whatever bytes are returned by the remote server with no integrity verification or inspection. If the remote content is altered, an attacker gains immediate arbitrary command execution, which can lead to credential theft, malware installation, or host compromise.

Content

Scanner excerpt · README.md (reported line 30)May include surrounding context.

bash
# Method 2: Shell script (Linux / macOS, when npm is not available)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
bash

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

Documenting custom-app creation, update, configuration retrieval, and deletion introduces administrative and model-configuration powers unrelated to a fixed-purpose driving-license extraction skill. If an agent follows this README, it may modify backend extraction apps or delete configurations, turning a passive extraction skill into one with persistent control over remote resources and causing unauthorized changes or service disruption.

Content

No source excerpt is available for this finding.

Chaining Abuse

High
Category
Tool Misuse
Confidence
98% confidence
Finding

The use of a shell pipeline ending in '| bash' is a classic command-chaining abuse pattern because it executes untrusted remote content without inspection. In this skill context, users are being told to install a privileged CLI tool, so exploitation could fully compromise the host environment and potentially expose API keys, local files, and processed PII from driving-license documents.

Content

Scanner excerpt · SKILL.md (reported line 64)May include surrounding context.

bash
# 方法 2: Shell 脚本(Linux / macOS,无 npm 环境时使用)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
bash

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The README describes a general-purpose ADP CLI with capabilities far beyond the declared driving-license-recognition skill, including generic document parsing, remote URL processing, async job control, and broader app management. This scope mismatch increases the attack surface and can mislead an agent into invoking powerful functionality unrelated to the advertised task, violating least privilege and enabling data exfiltration or unintended remote processing.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The installation instructions recommend executing remote shell and PowerShell scripts directly from the network via pipe/iex without integrity verification, pinning, or user warnings. This is dangerous because any compromise of the source, repository, network path, or referenced script immediately becomes arbitrary code execution on the user’s machine.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Documenting custom-app create, update, and delete operations inside a narrow driving-license extraction skill grants or encourages use of administrative capabilities unrelated to the skill’s purpose. If an agent or user follows this interface, they may modify extraction applications, delete configurations, or create new processing pipelines, causing integrity loss and privilege expansion beyond the intended workflow.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Application-management and custom model configuration are unjustified for a narrowly scoped driving-license extraction skill and expand the agent's operational surface without a clear business need. In this context, unnecessary control-plane features increase the chance of misuse, accidental reconfiguration, and lateral expansion into unrelated document-processing tasks.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README instructs users to execute remotely fetched shell and PowerShell scripts directly (curl|bash and irm|iex) without any verification or warning. This is dangerous because any compromise of the source repository, network path, or referenced script can immediately result in arbitrary code execution on the host where the skill is being installed or used.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill description and operational instructions are presented entirely in Chinese and describe the capability specifically for Chinese driving licenses, but they do not state that Chinese is required nor provide any user opt-in or alternative language path. Under the policy rule, forcing a specific language or locale without explicit choice or justified documentation is a natural-language policy issue.

Content

No source excerpt is available for this finding.

External Script Fetching

Low
Category
Supply Chain
Confidence
97% confidence
Finding

This command fetches a shell script from a remote URL and immediately executes it, creating a direct arbitrary-code-execution path. Because the skill is presented as operational guidance, users or agents may run it as-is, turning any upstream compromise into full command execution in the local environment.

Content

Scanner excerpt · README-CN.md (reported line 29)May include surrounding context.

bash
# 方法 2: Shell 脚本(Linux / macOS,无 npm 环境时使用)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
bash

External Script Fetching

Low
Category
Supply Chain
Confidence
93% confidence
Finding

The command fetches a shell script from a remote URL and executes it as part of installation, creating a direct trust dependency on mutable remote content. In an agent or automation environment, this becomes more dangerous because installation may be performed non-interactively, reducing opportunities for human review before code execution.

Content

Scanner excerpt · README.md (reported line 30)May include surrounding context.

bash
# Method 2: Shell script (Linux / macOS, when npm is not available)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
bash

External Script Fetching

Low
Category
Supply Chain
Confidence
97% confidence
Finding

The skill instructs users to fetch and execute a remote shell script directly from GitHub using curl piped to bash. This is dangerous because any compromise of the remote source, repository, network path, or referenced branch can lead to immediate arbitrary code execution on the user's system with the user's privileges.

Content

Scanner excerpt · SKILL.md (reported line 64)May include surrounding context.

bash
# 方法 2: Shell 脚本(Linux / macOS,无 npm 环境时使用)
curl -fsSL https://raw.githubusercontent.com/laiye-ai/adp-cli/main/scripts/adp-init.sh | bash
bash

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · license.md (reported line 10)May include surrounding context.

md
1. Licensing

1.1 Free Use and Distribution: The Licensor grants the Licensee a non-transferable, non-exclusive right to freely use, copy, publish, and distribute copies of the Product for non-commercial purposes. The aforementioned "non-commercial purposes" include, but are not limited to:
Personal learning, research, teaching, and evaluation.
Technical exchanges within academic institutions or open-source communities, non-profit projects.
Integration or demonstration in non-commercial products or services.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · license.md (reported line 14)May include surrounding context.

md
1. Licensing

1.1 Free Use and Distribution: The Licensor grants the Licensee a non-transferable, non-exclusive right to freely use, copy, publish, and distribute copies of the Product for non-commercial purposes. The aforementioned "non-commercial purposes" include, but are not limited to:
Personal learning, research, teaching, and evaluation.
Technical exchanges within academic institutions or open-source communities, non-profit projects.
Integration or demonstration in non-commercial products or services.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · license.md (reported line 22)May include surrounding context.

md
1. Licensing

1.1 Free Use and Distribution: The Licensor grants the Licensee a non-transferable, non-exclusive right to freely use, copy, publish, and distribute copies of the Product for non-commercial purposes. The aforementioned "non-commercial purposes" include, but are not limited to:
Personal learning, research, teaching, and evaluation.
Technical exchanges within academic institutions or open-source communities, non-profit projects.
Integration or demonstration in non-commercial products or services.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · license.md (reported line 32)May include surrounding context.

md
4. Disclaimer of Warranties

THE PRODUCT IS PROVIDED "AS IS", WITHOUT WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO: THE LICENSOR DOES NOT WARRANT THAT THE PRODUCT IS FREE FROM ERRORS, BUGS, WILL OPERATE PROPERLY, OR IS SUITABLE FOR A PARTICULAR PURPOSE; THE LICENSOR DOES NOT WARRANT THAT USE OF THE PRODUCT WILL NOT INFRINGE UPON THIRD-PARTY RIGHTS; THE LICENSOR SHALL NOT BE LIABLE TO THE LICENSEE OR ANY THIRD PARTY FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, OR PUNITIVE DAMAGES ARISING FROM THE USE OF THE PRODUCT.

5. Termination