The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interpolates the user-provided $ARGUMENTS directly into shell commands within the workflow and manual testing sections (e.g., ls -la $ARGUMENTS, {baseDir}/scanner.sh $ARGUMENTS, and apktool d -f -o ./decompiled $ARGUMENTS). If the input is not properly sanitized or escaped by the platform, this presents a risk of command injection where shell metacharacters in the path could execute arbitrary commands.
[DATA_EXFILTRATION]: By design, this skill extracts sensitive configuration data from provided APK files, such as API keys and Project IDs, and performs network requests to external Firebase endpoints (googleapis.com, firebaseio.com) to test for vulnerabilities. While these targets are well-known services and consistent with the skill's purpose, the process involves transmitting data extracted from the target files to remote servers.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (APK files) and provides detailed instructions for the agent to follow based on that data. While it includes defensive 'Rationalizations to Reject' to ensure the agent reports findings accurately, the ingestion of external data combined with the capability to execute shell commands and network requests creates an attack surface for indirect prompt injection.

firebase-apk-scanner