The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text content from external customer service cases and interaction logs. While its primary purpose is data aggregation and trend analysis, reading unsanitized user-supplied descriptions introduces a potential surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through the sn_customerservice_case, sys_journal_field, and interaction tables via the SN-Query-Table and SN-Natural-Language-Search tools described in SKILL.md.
Boundary markers: Absent; there are no specific encapsulation techniques or guidelines instructing the model to ignore potential instructions embedded within the customer cases.
Capability inventory: The execution context includes high-privilege capabilities such as SN-Execute-Background-Script and native Bash as defined in the SKILL.md manifest.
Sanitization: No escaping, content filtering, or input validation mechanisms are applied to the text fields before they are displayed or processed by the agent.

trending-topics