The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by ingesting and processing untrusted data from local files and user input. The lack of sanitization could allow malicious instructions stored in vaccination records to influence agent behavior.\n
Ingestion points: Data enters the agent context from data/allergies.json and data/vaccinations.json during the safety check and schedule generation phases (SKILL.md).\n
Boundary markers: There are no specific delimiters or instructions to ignore embedded commands within the processed data files.\n
Capability inventory: The skill uses the Write tool to modify data/vaccinations.json and the Read tool to access files, providing a persistent storage mechanism for injected instructions.\n
Sanitization: The skill does not implement escaping, validation, or filtering of the content retrieved from the data files before using it in the decision-making process.

vaccine