limrun-android-emulator
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
lim android adb-shellto execute commands directly on the Android emulator instance. While this is restricted to the emulator environment, it allows for arbitrary shell command execution as part of the tool's intended lifecycle management. - [EXTERNAL_DOWNLOADS]: The
lim android install-appcommand supports fetching and installing APK files from arbitrary URLs (e.g.,https://example.com/app.apk). This allows for the deployment of remote executable code onto the managed emulator instance. - [DYNAMIC_EXECUTION]: The skill interacts with the Android Debug Bridge (ADB) to perform low-level operations like injecting microphone audio (
play-on-microphone) and managing network bandwidth, which involves dynamic interaction with the emulator's system services. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the emulator that could be influenced by external applications.
- Ingestion points: UI hierarchy data via
element-tree(SKILL.md), application logs viaapp-log(SKILL.md), and visual state viascreenshot(SKILL.md). - Boundary markers: The skill does not define explicit boundary markers or delimiters when interpolating emulator data into the agent's context, although it does recommend using
grepto filterelement-treeoutput. - Capability inventory: The toolset includes powerful capabilities such as
adb-shell,push-file,install-app, andopen-url, which could be targeted by instructions hidden in processed UI text or log files. - Sanitization: There are no explicit instructions for the agent to sanitize or escape data retrieved from the emulator before processing it.
Audit Metadata