The Agent Skills Directory

[SAFE]: The skill is designed to manage medical consultations by reading data from the local data/ directory and writing reports. It includes explicit 'Safety Red Lines' that instruct the agent to avoid providing medication dosages, direct prescriptions, or definitive diagnoses. This demonstrates a strong adherence to safety best practices for medical-related AI agent skills.
[PROMPT_INJECTION]: The skill processes untrusted medical data from multiple files, which constitutes an indirect prompt injection surface. This risk is effectively mitigated by several design factors.
Ingestion points: Patient records, symptom logs, and chronic disease trackers located in the local data/ directory (e.g., data/health-feeling-logs.json).
Boundary markers: The subagent prompt templates in SKILL.md use clear markdown headers (## Skill 定义, ## 患者医疗数据) to separate instructions from untrusted patient data.
Capability inventory: The skill utilizes only Read and Write tools; it does not possess network access, shell execution capabilities, or administrative privileges.
Sanitization: While formal input sanitization is not specified, the agent's output is strictly constrained by safety protocols defined in the skill logic.

consult