The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill demonstrates patterns for ingesting untrusted data from external sources and user input which could be exploited in an indirect prompt injection attack.
Ingestion points: The WeatherApp example in references/examples.md ingests user input via the Input widget and external data from the wttr.in API.
Boundary markers: No explicit delimiters or instructions are provided to the model to treat the external API response as untrusted data.
Capability inventory: The skill examples use httpx for network requests and DirectoryTree for local file system access.
Sanitization: The fetch_weather method in references/examples.md updates the UI directly with the response.text from an external URL without sanitization.
[Data Exposure & Exfiltration] (LOW): Contains network operations targeting domains outside the predefined whitelist.
Evidence: references/examples.md contains a WeatherApp example that uses httpx to perform a GET request to https://wttr.in.

textual