android-design-explore

Warn

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The references/setup.md file describes a bootstrap process that automatically downloads an external JAR file (android-design-preview.jar) from GitHub Releases to the user's local cache directory.
  • [REMOTE_CODE_EXECUTION]: The skill operates by executing the downloaded android-design-preview.jar binary using Java. Running unverified external binaries presents a risk of executing arbitrary code on the host system.
  • [COMMAND_EXECUTION]: The skill frequently invokes shell commands via the android-design-preview tool (e.g., check, tree, open, install) to perform UI rendering and layout checks.
  • [INDIRECT_PROMPT_INJECTION]: The skill reads several files from the local repository that may contain untrusted content.
  • Ingestion points: The agent is instructed to read DESIGN.md, design/*.explore.kt, and design/assets/* as the basis for its design tasks.
  • Boundary markers: There are no specific instructions or delimiters provided to prevent the agent from following malicious instructions that might be embedded in these files.
  • Capability inventory: The skill possesses the ability to execute shell commands and modify files (e.g., updating DESIGN.md), which could be leveraged if an injection attack occurs.
  • Sanitization: Content from repository files is processed without explicit sanitization or validation steps.
  • [PRIVILEGE_ESCALATION]: In references/setup.md, the skill explicitly instructs users in restricted environments like Codex to modify their configuration (~/.codex/config.toml) to enable network_access = true and add writable_roots. This guidance encourages users to weaken the security posture of their execution environment.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 20, 2026, 03:10 AM