The Agent Skills Directory

Untrusted Data Ingestion (Indirect Prompt Injection): The skill is designed to process external input from clients during the intake process. While processing untrusted text is a necessary part of the skill's primary function, it technically introduces a surface where a client might attempt to influence the agent's behavior or conclusions via their narrative. The skill mitigates this by including clear 'AI-ASSISTED DRAFT' labels, providing specific verification prompts for the student to confirm dates and facts, and explicitly stating that all summaries require professional legal review and attorney oversight.
Access to Configuration Files: The skill reads operational parameters (practice areas, templates, supervision style) from specific local paths within the agent's configuration directory. This is standard behavior for defining a skill's execution environment and clinic-specific rules, ensuring the agent operates within the intended legal context.

client-intake