The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents methods for ingesting untrusted data through serverless function handlers (context.Req.Body, context.Req.BodyJson) and database operations. While this is a standard feature, it creates a surface where external inputs could contain malicious instructions meant for downstream agent processing.
Ingestion points: Appwrite Function request bodies in src/Main.cs and database row content in TablesDB.CreateRow.
Boundary markers: None explicitly mentioned in the code examples for separating data from instructions.
Capability inventory: File system access via Storage, database modifications via TablesDB, and serverless execution via Functions.
Sanitization: No explicit sanitization or validation logic is shown in the examples when handling request payloads or database fields.
[CREDENTIALS_UNSAFE]: The skill correctly demonstrates secure secret management by retrieving API keys and project IDs from environment variables (Environment.GetEnvironmentVariable) rather than hardcoding them.
[EXTERNAL_DOWNLOADS]: The skill references the official Appwrite package from the NuGet registry and connects to the official vendor domain appwrite.io. These are legitimate resources owned by the skill's author.

appwrite-dotnet