The Agent Skills Directory

[SAFE]: The skill demonstrates standard network interaction patterns using Godot's built-in HTTPRequest and WebSocketPeer classes. All examples use placeholder URLs (e.g., api.example.com) and provide sound security advice, such as avoiding the inclusion of secrets in client builds and using HTTPS/WSS.
[EXTERNAL_DOWNLOADS]: Includes recipes for downloading external content directly to the user:// directory. Evidence: http.download_file = path in references/downloads-and-uploads.md. This is a standard feature for fetching game assets.
[INDIRECT_PROMPT_INJECTION]: The skill identifies surfaces for untrusted data ingestion.
Ingestion points: Data is received via HTTPRequest.request_completed in SKILL.md and references/rest-client-patterns.md, and through WebSocketPeer.get_packet() in SKILL.md.
Boundary markers: No explicit instructions to ignore embedded commands within the fetched data are provided.
Capability inventory: The skill has capabilities for network communication and file writing (DirAccess.rename_absolute()).
Sanitization: External data is parsed using JSON.parse_string(), which is a safe parsing method but does not sanitize the logical content for downstream AI consumers.
[DYNAMIC_EXECUTION]: Mentions the use of ProjectSettings.load_resource_pack() to mount downloaded content packs. While this allows for dynamic loading of game resources, it is a standard engine capability.

web-services