The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes market titles, descriptions, and rules directly from the Kalshi public API. This creates an attack surface where an adversary controlling content on the prediction market could attempt to influence the agent's behavior.
Ingestion points: Multiple scripts, including scripts/markets.py and scripts/search.py, fetch external content from https://api.elections.kalshi.com.
Boundary markers: The skill instructions do not currently include explicit boundary markers or directions for the agent to ignore instructions embedded within the market data.
Capability inventory: The skill has the capability to write to the local filesystem (for caching in scripts/search.py) and perform network operations to the Kalshi API.
Sanitization: Data retrieved from the API is displayed or processed without specific sanitization for potential injection patterns.
[EXTERNAL_DOWNLOADS]: The skill scripts fetch prediction market data and exchange status from Kalshi's official API endpoints.
Evidence: Scripts connect to https://api.elections.kalshi.com/trade-api/v2 to retrieve market-related information.

kalshi-markets