The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a hierarchy for real-time data that processes input from external sources (via stream::set, stream::update, and stream::send) to trigger reactive functions. This architecture allows untrusted data to potentially influence the logic of downstream handlers.
Ingestion points: Data enters the system through the stream::set, stream::update, and stream::send functions described in SKILL.md.
Boundary markers: The instructions do not specify any boundary markers, delimiters, or sanitization requirements for the data payloads within the streams.
Capability inventory: The skill utilizes iii.registerTrigger to invoke other functions (function_id) across the platform, creating a chain where stream data can drive broader system actions.
Sanitization: There is no evidence of data validation, filtering, or escaping mentioned in the skill documentation to mitigate malicious content in the item payloads.

iii-stream