The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted data from external sources.
Ingestion points: The skill reads from sn_customerservice_case (description), sys_journal_field (comments), and sys_email (email bodies) which often contain content directly supplied by external customers.
Boundary markers: The skill lacks explicit instructions for the agent to ignore or delimit potentially malicious instructions embedded within the retrieved customer data.
Capability inventory: No dangerous execution capabilities (like arbitrary command execution or local file writing) were identified in the provided scripts.
Sanitization: There is no evidence of sanitization or filtering of retrieved data before it is presented to the model for summarization.

resolution-notes