The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted market descriptions from the Novig exchange, which could contain malicious instructions designed to influence agent behavior.
Ingestion points: Data enters the context via get_markets, get_event, and search_markets commands, specifically in the market descriptions and _raw provider payloads.
Boundary markers: The instructions lack specific delimiters or boundary markers to isolate untrusted text, though they do provide a textual warning to the agent to treat descriptions as untrusted.
Capability inventory: The skill uses the requests library for data retrieval but is limited to read-only operations, reducing the potential impact of an injection.
Sanitization: No specific sanitization or filtering of the remote content is mentioned before it is processed by the agent.

novig