The Agent Skills Directory

[CREDENTIALS_UNSAFE]: All scripts in the repository (e.g., scripts/maps_directions.py, scripts/maps_geocode.py, etc.) contain a hardcoded fallback API key BETA_TEST_KEY_MARCH_2026. While labeled as a test key, hardcoding credentials in source code is a security risk as it can lead to unauthorized use of the associated service account.
[PROMPT_INJECTION]: The skill processes untrusted user input from command-line arguments (such as --address, --query, --origin, and --destination) and passes them directly to the OneKeyAgentRouter. This creates a surface for indirect prompt injection where an attacker could provide data that influences the behavior of the AI agent or the external API router.
Ingestion points: User-controlled parameters in scripts/maps_geocode.py (address), scripts/maps_search_places.py (query), scripts/maps_distance_matrix.py (origins/destinations), and scripts/maps_directions.py (origin/destination).
Boundary markers: No delimiters or safety instructions are used to wrap the user input before it is passed to the router.
Capability inventory: The skill performs network operations via the OneKeyAgentRouter.invoke method to access Google Maps APIs.
Sanitization: The input values are parsed for type correctness (e.g., float for coordinates) but are not sanitized for potential prompt injection or escape sequences.

google-maps