software-android-native

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to use standard Android command-line tools, including the Gradle wrapper (./gradlew) and Android Debug Bridge (adb). These commands are used for legitimate development tasks such as building APKs, installing applications, launching activities, and inspecting logs. Instructions include best practices such as verifying package names and ensuring the emulator is available before execution.
  • [EXTERNAL_DOWNLOADS]: The skill references multiple external documentation sources from trusted organizations. These include official Google Android documentation (developer.android.com), Kotlin language references (kotlinlang.org), and official agent platform documentation from OpenAI and Anthropic. These resources are used neutrally to provide technical context and compliance guidance for Android development and store policies.
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes workflows that ingest external data, specifically Android project source code and UI hierarchy dumps generated by uiautomator. While this ingestion creates a surface for indirect prompt injection, the skill mitigates this through a rigorous 'proof-first' execution loop. This requires the agent to validate all changes through successful builds, targeted automated tests (JUnit 5, Compose Testing), and visual verification via screenshots before completing a task.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 02:03 PM