kalshi
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION] (LOW): The skill provides examples using
curlandjqto fetch and parse JSON data from prediction markets. This is standard for CLI-based data tools. - [EXTERNAL_DOWNLOADS] (LOW): The skill makes network requests to
api.elections.kalshi.comandgamma-api.polymarket.com. These are the correct endpoints for the described service but are not on the pre-approved whitelist. - [PROMPT_INJECTION] (LOW): Categorized as a potential Indirect Prompt Injection surface. 1. Ingestion points: Data is pulled from external prediction market APIs. 2. Boundary markers: None. 3. Capability inventory: Commands are limited to data fetching and parsing. 4. Sanitization: Structural validation is performed by
jqduring the parsing process, which limits the risk of unstructured command injection.
Audit Metadata