core-nfc
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements logic to ingest and process untrusted NDEF data from external physical tags, creating an entry point for untrusted input.\n
- Ingestion points: Untrusted data enters the application context via
readerSession(_:didDetectNDEFs:)inSKILL.mdandreferences/nfc-patterns.md, and throughNSUserActivity.ndefMessagePayloadinSKILL.md.\n - Boundary markers: The provided Swift implementation patterns do not include delimiters or specific instructions for the agent to treat the NDEF payload as untrusted data or to disregard instructions found within the data records.\n
- Capability inventory: The skill includes code for scanning, connecting, reading, writing, and permanently locking NFC tags, but does not include network or local file system write operations in the provided snippets that would significantly escalate the risk.\n
- Sanitization: NDEF payloads are decoded directly to strings (e.g., using
String(decoding:record.payload, as: UTF8.self)) without content validation, filtering, or sanitization, making the resulting strings a vector for injection if passed back to an AI model.
Audit Metadata