The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a monitoring framework for clinical queries, which could be exploited if malicious instructions are embedded within the queries being observed by the agent.
Ingestion points: The skill monitors clinical queries via the trackClinicalQuery instrumentation defined in SKILL.md.
Boundary markers: The instructions include clear warnings to avoid logging patient identifiers or query text to maintain HIPAA compliance, serving as a data boundary.
Capability inventory: The skill configuration allows for Read, Write, Edit, and Grep tools, which could be leveraged if an injection is successful.
Sanitization: The references/implementation.md file specifies the use of pino logger redaction for sensitive fields such as patientId, mrn, and ssn.

openevidence-observability