arbitrage
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFE
Full Analysis
- [Prompt Injection] (SAFE): No instructions attempting to override agent behavior or bypass safety filters were identified.
- [Data Exposure & Exfiltration] (SAFE): The skill uses environment variables for API keys (POLY_API_KEY, KALSHI_API_KEY) which is standard practice. No sensitive local file access or unauthorized network exfiltration patterns were found.
- [Unverifiable Dependencies] (SAFE): The documentation references an external library 'clodds/arbitrage' in code examples. However, no suspicious installation commands or remote execution patterns (curl|bash) were detected.
- [Indirect Prompt Injection] (SAFE): The skill ingests market data such as question titles and IDs. While this represents a data ingestion surface, the skill lacks high-privilege capabilities (like shell execution or file system writing) that would allow for exploitation via poisoned market data.
- [Obfuscation] (SAFE): No encoded strings, hidden characters, or homoglyphs were found in the skill file.
Audit Metadata