The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from smart home accessories, including accessory names, service types, and characteristic values, which are user-controlled and could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: accessory.name, service.serviceType, and characteristic.value are read in code snippets within SKILL.md and references/matter-commissioning.md.
Boundary markers: The skill lacks explicit instructions or delimiters to ensure the agent ignores or sanitizes instructions that might be embedded in these external data points.
Capability inventory: The skill facilitates writing characteristic values and commissioning new Matter devices, which could be triggered maliciously if the agent is influenced by poisoned accessory metadata.
Sanitization: There is no evidence of validation or sanitization for the environment-sourced strings before they are processed by the agent.

homekit