skills/iskenkenya/skills/android-cli/Gen Agent Trust Hub

android-cli

Fail

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download CLI tools and installation scripts from dl.google.com. These downloads are part of the standard deployment of the Android development environment.\n- [REMOTE_CODE_EXECUTION]: Installation commands for Linux and macOS pipe remote scripts directly into bash, while the Windows command executes a downloaded CMD file. These operations target official Google servers.\n- [COMMAND_EXECUTION]: The skill utilizes the android CLI and adb (Android Debug Bridge) to interact with physical or virtual devices, manage software packages, and execute shell-level input commands.\n- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it evaluates data from external or untrusted sources to determine agent actions.\n
  • Ingestion points: Processes <action> elements from XML journey files (references/journeys.md) and extracts UI attributes like text and contentDesc from device layout dumps (references/interact.md).\n
  • Boundary markers: The skill instructions do not establish explicit boundary markers or warnings to the agent regarding potential instructions embedded within the processed XML or JSON data.\n
  • Capability inventory: The agent can execute a wide range of android and adb shell input commands based on the strings it reads from the device or test files.\n
  • Sanitization: No evidence is provided that the skill performs validation or sanitization of input strings before incorporating them into command execution or decision-making logic.
Recommendations
  • HIGH: Downloads and executes remote code from: https://dl.google.com/android/cli/latest/darwin_x86_64/install.sh, https://dl.google.com/android/cli/latest/linux_x86_64/install.sh, https://dl.google.com/android/cli/latest/darwin_arm64/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 22, 2026, 10:16 AM
Security Audit — agent-trust-hub — android-cli