bybit-v5
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. Ingestion points include user-provided credential files and Bybit API response JSON. Boundary markers are absent for user-provided file ingestion. Capability inventory includes network requests via curl equivalents and file writing to TOOLS.md. Sanitization includes instructions to mask secrets and require "CONFIRM" for transactions. Metadata lists 'Bybit' as author, which differs from the author context 'johnnywic', though common for API wrappers.
- [DATA_EXFILTRATION]: The skill handles sensitive financial credentials and performs transaction operations. It implements safety mitigations such as masking API keys (first 5 and last 4 characters) and secrets (last 5 characters only). It also requires a manual confirmation step for all POST requests to mainnet trading endpoints.
- [COMMAND_EXECUTION]: Documentation includes shell command examples using openssl for HMAC-SHA256 signing and curl for interacting with the Bybit API. These utilities are standard for the skill's intended purpose of secure API communication.
Audit Metadata