The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill's workflow encourages the agent to search for and read existing code within the workspace (using tools like rg) to identify local design patterns. Because the agent processes this untrusted project data and subsequently has the capability to write new files and execute build/verification commands, there is a potential surface for indirect prompt injection if the workspace contains code designed to influence the agent's behavior.
Ingestion points: Workspace files accessed during pattern discovery (e.g., SKILL.md instructions to use rg).
Boundary markers: Not explicitly defined in the instructions.
Capability inventory: The skill assumes the ability to modify project files and execute build tools to verify code correctness (e.g., SKILL.md workflow).
Sanitization: Not present in the provided instructions.
[External Repository Reference]: The documentation suggests the use of a specific third-party library for navigation logic. While presented as a recommendation, this library originates from a source outside of the recognized trusted organizations, and its inclusion in a project represents an external dependency that should be evaluated by the developer.
Evidence: Found in references/navigationstack.md.
[Information Disclosure]: The reference files contain hardcoded absolute paths that appear to originate from the development environment used during the skill's creation. These paths disclose information about internal directory structures and local usernames.
Evidence: Absolute paths found in references/scroll-reveal.md.

swiftui-ui-patterns