The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documentation describes how to set up a client that consumes data from a remote producer, creating a potential attack surface for indirect prompt injection.
Ingestion points: The on_message callback in Python and the C message callback register handlers for text data sent from the remote producer. Additionally, wait_frame ingests encoded video data from the network.
Boundary markers: The provided documentation and code snippets do not include or recommend the use of delimiters or instructions to ignore potential commands embedded in the incoming data stream.
Capability inventory: The skill has access to Read and Grep tools and provides networking capabilities to send messages and input events back to the producer.
Sanitization: There is no evidence of data sanitization, schema validation, or escaping of the incoming text messages before they are processed or printed to the output.

native-consumers