The Agent Skills Directory

[METADATA_POISONING]: The skill's metadata and instructional content reference a future operating system version (iOS/iPadOS 26, macOS 26) that does not exist. This misleading information could cause an AI agent to generate code for non-existent APIs or make incorrect assumptions about the runtime environment.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates how to extract and process indexable text from PaperMarkup documents, creating a vulnerability surface where instructions embedded in document content could influence agent behavior. * Ingestion points: PaperMarkup data is loaded from external files via PaperMarkup(dataRepresentation: data) in references/paperkit-patterns.md. * Boundary markers: The instructions do not define delimiters or specific warnings to ignore instructions embedded within the document's text boxes or shapes. * Capability inventory: The skill patterns include local file system access (data.write) and UI presentation/delegation. * Sanitization: There is no mention of sanitizing or validating the indexableContent extracted from documents before it is processed or indexed.
[EXTERNAL_DOWNLOADS]: The skill includes documentation links to sosumi.ai (e.g., https://sosumi.ai/documentation/paperkit). This is an external third-party domain not associated with the official documentation for the framework.

paperkit