The Agent Skills Directory

[PROMPT_INJECTION]: The FAQ component in references/faq-howto.md renders FAQ answers using the set:html directive. This bypasses Astro's default HTML escaping and creates a surface for indirect prompt injection if the source data for the FAQ is attacker-controlled.\n
Ingestion points: The faqs property in the FAQ component and the associated FAQSchema.astro component.\n
Boundary markers: Absent; no delimiters or instructions are provided to the agent to ignore embedded instructions within the FAQ content.\n
Capability inventory: No dangerous runtime capabilities (e.g., subprocess execution, filesystem writes) were identified in the provided skill code.\n
Sanitization: Absent; the code explicitly uses set:html for raw rendering of strings.

schema-patterns