obtain-screenshot-android

Fail

Audited by Gen Agent Trust Hub on Feb 13, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis

The skill instructs the execution of adb commands, which are powerful command-line utilities for interacting with Android devices. It also instructs the execution of ./gradlew installDebug. The gradlew script is a wrapper for Gradle, a build automation tool. Executing ./gradlew can trigger arbitrary code execution defined in the project's build.gradle files, download external dependencies, and perform various system operations. The content of these Gradle files is not provided for analysis, making this a significant unverifiable dependency and command execution risk. The skill uses placeholders like <device_serial>, <package>, and <preview-component-package-name> in shell commands. If the AI agent substitutes these placeholders with user-provided input without proper sanitization, it could lead to command injection vulnerabilities, allowing an attacker to execute arbitrary shell commands on the host system. The README.md mentions kimi-cli from MoonshotAI/kimi-cli. MoonshotAI is not on the list of trusted GitHub organizations, making this an unverifiable external dependency. The skill writes screenshot files (.png) to the local filesystem. While the skill itself doesn't exfiltrate these, their presence on the local machine could pose a data exfiltration risk if they contain sensitive information and are later accessed by other malicious processes or commands.

Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 13, 2026, 07:19 AM