The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bundled shell script (scanner.sh) to execute local security tools including apktool for decompilation and curl for verifying endpoint security. These operations are essential to its function as a vulnerability scanner.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted third-party data from Android APK files (smali code, XML resources, and binary strings). This represents a vulnerability surface, but the risk is mitigated by the disable-model-invocation: true configuration, which prevents the model from interpreting extracted data as agent instructions.
[DATA_EXFILTRATION]: Network operations are strictly limited to official Google and Firebase domains (googleapis.com, firebaseio.com, etc.) for the purpose of testing the security of the target application's backend. No unauthorized exfiltration of local credentials or sensitive files was detected.
[SAFE]: The skill demonstrates high-quality security hygiene, including explicit authorization warnings, automated cleanup of test data, and detailed remediation guidance for findings. The source code is transparently documented and adheres to standard security research practices.

firebase-apk-scanner