The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for processing data records fetched from CloudKit, which represents an ingestion point for untrusted external content.
Ingestion points: Data is ingested through operation.recordWasChangedBlock in references/cloudkit-patterns.md and changes.modifications in references/core-implementation.md.
Boundary markers: No delimiters or warnings are specified to prevent the agent from treating data within records as instructions.
Capability inventory: The skill scripts have the capability to write to the CloudKit database (privateDB.save), write state to the local file system (data.write), and move files (FileManager.default.moveItem).
Sanitization: The instructions do not include steps for sanitizing or validating the content of external records before processing.
[EXTERNAL_DOWNLOADS]: The skill includes reference links to sosumi.ai, which is a non-standard third-party domain for documentation.
Evidence: Several links in SKILL.md point to https://sosumi.ai/documentation/... instead of official documentation sites.

cloudkit