The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions define a processing pipeline for untrusted data, creating a surface for indirect prompt injection.
Ingestion points: The skill processes input from $ARGUMENTS and content retrieved via external tools such as WebFetch (referenced in Step 3 of SKILL.md).
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions within the processed data.
Capability inventory: The skill references external tools for web retrieval but does not include its own executable scripts or complex tool definitions.
Sanitization: No validation or sanitization of the content retrieved from external URLs is specified before the AI incorporates it into its response.

cite-sources