The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes an environment for processing untrusted user data, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through fields in the claim_data dictionary (SKILL.md) and the application dictionary (references/POLICY_ADMINISTRATION_SYSTEM.md).
Boundary markers: The skill lacks explicit boundary markers or instructions to treat embedded text in data as non-executable instructions.
Capability inventory: The skill is configured with Read, Write, and Edit tools, allowing for the manipulation of files and data structures based on processed information.
Sanitization: The provided logic does not include sanitization or strict validation of user-input strings before processing.

insurance-expert