The Agent Skills Directory

[METADATA_POISONING]: The skill attributes itself to 'Paul Hudson' (a prominent Swift educator) in the YAML metadata, which conflicts with the actual author account 'jordancoin'. This mismatch is deceptive and misrepresents the source and validation of the content.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and review user-provided Swift source code, which establishes a vulnerability surface for indirect prompt injection. 1. Ingestion points: Swift and SwiftUI source files provided for code review. 2. Boundary markers: The skill does not define clear delimiters or use 'ignore embedded instructions' warnings for the code being reviewed. 3. Capability inventory: The skill is capable of summarizing findings, suggesting code changes, and potentially utilizing tools like RenderPreview (mentioned in references/hygiene.md) if an Xcode MCP is available. 4. Sanitization: No sanitization or validation of the input code is specified.
[PROMPT_INJECTION]: The skill includes instructions that force a specific, non-factual context ('iOS 26 exists, and is the default deployment target'). This 'pretend' injection technique is used to override the agent's factual grounding to force specific behavioral outputs.

swiftui-pro