The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses the WebFetch tool to retrieve content from URLs provided by the user. This is a primary function of the skill and targets user-specified locations. The eval example explicitly references raw.githubusercontent.com/anthropics/, which is a trusted source.
[COMMAND_EXECUTION]: Uses Glob or Bash ls to list folder contents for the purpose of creating a file loading checklist. This is routine behavior for directory management and includes sensible filtering to skip sensitive or hidden directories (e.g., .git/, .env).
[INDIRECT_PROMPT_INJECTION]: As the skill is designed to ingest untrusted data from the web or local files into the context window, it possesses a vulnerability surface for indirect prompt injection. However, it does not perform any high-risk automated actions with that data in the same turn; it simply reads the content and stops (responding 'DONE').
[SAFE]: No malicious patterns, obfuscation, or persistence mechanisms were detected. The skill's behavior matches its stated purpose.

load-docs