The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements functionality to ingest and process data from external NFC tags, which are untrusted input sources. This creates a potential attack surface where instructions embedded in tag payloads could influence agent behavior.
Ingestion points: Data is read through NFCNDEFReaderSessionDelegate and NFCTagReaderSessionDelegate methods in SKILL.md and references/nfc-patterns.md, as well as via NSUserActivity for background reading.
Boundary markers: The code snippets do not include delimiters or specific instructions to the AI to ignore embedded commands within the tag data.
Capability inventory: The skill allows writing NDEF messages to tags and displaying user alerts, which could be misused if the agent is tricked by malicious tag content.
Sanitization: There is no evidence of validation or sanitization of the strings, URLs, or custom payloads extracted from the tags.
[EXTERNAL_DOWNLOADS]: The skill includes multiple references to an external, non-standard domain for technical documentation.
Evidence: The 'References' section in SKILL.md links to https://sosumi.ai/documentation/ for various CoreNFC components. While these appear to be informational links rather than direct script downloads, they point to a third-party domain that is not recognized as a trusted industry source.

core-nfc