The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates processing messages from WebSocket channels, which creates a surface where instructions from external sources can enter the agent's context.
Ingestion points: Data enters via the blink.realtime.subscribe and channel.onMessage handlers described in SKILL.md.
Boundary markers: The documentation does not specify the use of delimiters or 'ignore' instructions to separate incoming message data from system instructions.
Capability inventory: The skill includes the blink_realtime_publish tool, which allows the agent to write back to channels based on processed input.
Sanitization: No sanitization, validation, or filtering of the incoming data object is mentioned in the provided examples.

blink-realtime