The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external information such as compiler diagnostics and third-party documentation links (e.g., from sosumi.ai), which could potentially serve as a vector for malicious instructions. This risk is mitigated by explicit instructions to prioritize primary official sources like Swift Evolution and official SDK headers for all technical verification.
[SAFE]: The skill references the apple/swift-async-algorithms repository, which is a trusted source for asynchronous sequence utilities.
[SAFE]: The instructions for using manual synchronization opt-outs (nonisolated(unsafe) and @unchecked Sendable) include mandatory requirements for documented safety proofs, ensuring that memory safety invariants are maintained when bypassing compiler checks.

swift-concurrency