The Agent Skills Directory

[Indirect Prompt Injection] (INFO): The skill includes an LLM-as-judge protocol that processes untrusted model responses. Ingestion points: prompt templates in SKILL.md. Boundary markers: Absent. Capability inventory: Display and reasoning only (Tier: INFO). Sanitization: None specified.
[Data Exposure & Exfiltration] (SAFE): No hardcoded secrets, sensitive file paths, or network exfiltration patterns were detected in the skill content.
[Remote Code Execution] (SAFE): The skill consists of documentation and code templates; it does not contain executable scripts or external dependency requirements.

Model Evaluator