The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process untrusted text data from medical messages, creating a surface for indirect prompt injection. An attacker could attempt to manipulate the triage outcome by embedding instructions within the message body.
Ingestion points: subject and body fields in the input message objects.
Boundary markers: None specified to differentiate between data and instructions.
Capability inventory: No executable tools or scripts are defined within the skill.
Sanitization: No input validation or sanitization mechanisms are described.

medical-triage