The Agent Skills Directory

[SAFE]: The skill provides instructional content and code examples for desktop application development using the GPUI framework. All external resources, such as the gpui-component crate, are consistent with the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill provides templates for building interactive UI components (e.g., the Markdown editor with slash commands in examples/markdown_editor.rs) that ingest untrusted user text. While this creates a potential surface for indirect prompt injection in applications built using these templates, the implementation is standard for UI frameworks.
Ingestion points: Untrusted data enters the context via the Input component and the CompletionProvider trait (e.g., in reference/input-lsp.md and examples/markdown_editor.rs).
Boundary markers: None are present in the code snippets to delimit untrusted input from agent instructions.
Capability inventory: The generated code focuses on UI rendering and state management; no dangerous file system or network operations are included in the provided snippets.
Sanitization: No input sanitization or validation logic is shown in the provided templates.

gpui-components