The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process local metadata files (e.g., titles, descriptions, and assets) from the fastlane/metadata/android directory. While this is the intended functionality, it creates an attack surface where malicious content inside those files could influence agent behavior during a sync or validation workflow.
Ingestion points: Metadata files within fastlane/metadata/android read by gpd migrate commands.
Boundary markers: No specific delimiters or safety warnings for the agent regarding the content of these files are present in the instructions.
Capability inventory: The skill uses the gpd CLI to perform network operations (publishing to Google Play) and file system operations (reading/writing metadata and assets).
Sanitization: The instructions do not mention sanitizing the input data, though they suggest using gpd migrate fastlane validate to catch schema errors.
[Command Execution] (SAFE): All commands use the gpd utility with standard arguments. There are no signs of command injection or attempts to execute arbitrary system shells.
[Data Exposure] (SAFE): No hardcoded API keys, tokens, or sensitive file paths (like SSH keys) are present. It uses generic placeholders like com.example.app.
[Prompt Injection] (SAFE): No override instructions, bypass markers, or instructions to ignore system safety protocols were detected.

gpd-metadata-sync