swift-patterns
Warn
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill contains deceptive technical documentation in
references/modern-swiftui-apis.md. It presents fabricated APIs for "iOS 26" and references future events (WWDC25) as factual sources. This misleading information could cause an AI agent to suggest non-existent or non-functional code patterns to users. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user-provided SwiftUI code for review and refactoring, creating a surface for indirect prompt injection.
- Ingestion points: User-provided Swift and SwiftUI source code files passed to the agent for review or refactoring workflows described in
SKILL.md. - Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the user-provided code comments or strings.
- Capability inventory: The skill encourages the agent to perform code refactoring, which typically involves file system write operations.
- Sanitization: There is no mention of sanitization or validation of the input code to prevent embedded instructions from influencing the agent's behavior.
Audit Metadata