The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of adb shell commands, including root commands (su -c), to interact with the Android operating system, pull APK files, and access protected application data in /data/data/. This is necessary for deep inspection of mobile applications.
[EXTERNAL_DOWNLOADS]: The documentation in references/frida.md recommends the use of Frida's --codeshare feature to download and execute community-contributed scripts for SSL unpinning and root detection bypass. While these are remote scripts from third-party contributors, they are standard resources in the security community.
[DATA_EXFILTRATION]: By design, the skill extracts and logs sensitive information from target applications, including PII, authentication tokens (JWTs), and credentials. These are stored in a local session directory (e.g., /tmp/workers-app-tester-...). The scripts/capture.py script specifically includes a PRESERVE_AUTH flag to log full authentication headers for testing purposes.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. It ingests data from untrusted target applications (via UI hierarchy parsing in scripts/ui.py and network traffic in scripts/traffic.py). If a target app contains malicious instructions in UI labels or response bodies, the agent might attempt to follow them. The skill's 'Rules' provide some operational guardrails to mitigate this risk.

workers-app-tester