The Agent Skills Directory

[SAFE]: The skill consists of high-quality technical documentation for Swift Concurrency. Analysis of the scripts and instruction files shows no evidence of prompt injection, credential exfiltration, or malicious persistence mechanisms. The content is aligned with professional software engineering best practices.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest project configuration data from files such as Package.swift and project.pbxproj to determine the appropriate concurrency behavior (e.g., Swift 6 mode, strict concurrency checking). While this establishes a surface for indirect prompt injection via untrusted content in local project files, the associated risk is judged to be negligible given the agent's limited capability profile within this skill (advisory only).
Ingestion points: SKILL.md (instructions to read Package.swift and project.pbxproj).
Boundary markers: Absent for local file ingestion.
Capability inventory: The skill is restricted to providing technical advice; it does not execute subprocesses, perform network exfiltration, or modify system files based on the analyzed project data.
Sanitization: Content read from local project files is not explicitly sanitized by the skill's instructions.

swift-concurrency