The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for ingesting external data through APIs such as MKLocalSearch and CLGeocoder. While these represent an attack surface for indirect prompt injection if user-provided strings are processed without sanitization, the skill focuses on technical implementation and basic data validation. * Ingestion points: Data returned from Apple's location services (MKLocalSearch, CLGeocoder, MKLocalSearchCompleter) and user-provided search queries interpolated into API calls. * Boundary markers: The suggested code patterns do not include explicit prompt boundary markers or instruction-ignoring delimiters for external data. * Capability inventory: The skill includes code for asynchronous network requests, geocoding, and map rendering, but does not involve direct file-system writes or shell command execution based on untrusted data. * Sanitization: The skill recommends validating coordinates and filtering invalid locations (e.g., horizontalAccuracy >= 0), which mitigates data-integrity issues but does not address prompt-level sanitization.

axiom-location