The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs Glob, Grep, and Read tools to scan the user's project files for existing state management and navigation patterns. This ingestion of untrusted external data to guide its code generation logic creates an indirect prompt injection surface.
Ingestion points: Scans project files such as **/*StateRestoration*.swift and **/*AppState*.swift (SKILL.md).
Boundary markers: None are present to distinguish between the content of the read files and the agent's instructions.
Capability inventory: The skill has access to Write, Edit, and Bash tools, which allow for filesystem modification and shell command execution (SKILL.md).
Sanitization: There is no mention of filtering or sanitizing the content ingested from the project files before it is processed by the agent.
[SAFE]: The Swift code templates provided in templates.md for AppState.swift, StateRestorationManager.swift, and other components follow official platform conventions for UI state persistence using Codable and @Observable without any malicious functionality.
[SAFE]: The skill includes explicit security warnings under the 'Gotchas' section, advising developers to never persist passwords, tokens, or PII in state restoration files and to use the Keychain for sensitive data instead.

state-restoration