The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network requests to fetch mapping data from established public mapping and timezone services.\n
Evidence: maps_client.py connects to nominatim.openstreetmap.org, overpass-api.de, router.project-osrm.org, and timeapi.io to retrieve location, routing, and timezone information. These are recognized mapping infrastructure providers and no sensitive local data is transmitted during these operations.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied location names and retrieves data from community-maintained sources (OpenStreetMap), which creates a surface for potential indirect prompt injection.\n
Ingestion points: User input for place names, addresses, and landmarks in the search, nearby, distance, directions, and area commands in SKILL.md.\n
Boundary markers: The instructions provided to the agent do not currently include boundary markers or explicit directions to ignore instructions potentially embedded in the mapping data results.\n
Capability inventory: The skill performs network requests to fetch location data and returns processed information, including coordinates and navigation URLs, to the agent.\n
Sanitization: While the script uses standard URL encoding for outbound API requests, it does not include specific logic to sanitize retrieved map feature data for malicious instructions before it is presented to the agent.

maps