The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from an external source (emails) which could contain malicious instructions.
Ingestion points: The spark thread <id> command mentioned in SKILL.md reads the full content of email threads into the agent's context.
Boundary markers: The recipe does not include explicit instructions or markers to treat the email content as data rather than instructions.
Capability inventory: The skill utilizes the spark CLI tool to perform actions such as archiving, moving to folders, pinning, snoozing, and drafting replies.
Sanitization: No sanitization or validation of the email body is performed before the agent reviews and acts on the thread.

spark-recipe-inbox-zero