crypto-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external API endpoints (Binance), which constitutes a potential surface for indirect prompt injection if the remote content were manipulated.\n
- Ingestion points: Market data, funding rates, and open interest are fetched from
fapi.binance.comas described inreferences/api-endpoints.md.\n - Boundary markers: The instructions do not define specific delimiters for the API response, but the agent is expected to parse structured JSON data.\n
- Capability inventory: The agent is restricted to network operations (
curl) for data retrieval. No filesystem modification or system-level execution capabilities are requested.\n - Sanitization: While no explicit sanitization logic is provided in the instructions, the data consists of financial metrics (numbers and short strings) which have a low risk of containing malicious instructional content.\n- [SAFE]: The skill targets well-known and reputable API services (
binance.com) for its core functionality. It does not access sensitive local files, use obfuscated code, or attempt to persist on the host system. The instructions for periodic execution (cron) are consistent with platform-supported scheduled tasks for automated reporting.
Audit Metadata