fiat
Fail
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructs the agent to search for and extract BINANCE_API_KEY and BINANCE_SECRET_KEY from environment variables and specific files on the local system.
- Evidence: Found in references/sapi-endpoints.md under the "Security" section.
- [DATA_EXFILTRATION]: Instructions direct the agent to automatically scan sensitive file paths including ~/.env, ~/.openclaw/secrets.env, and .env in the workspace to find credentials.
- Evidence: Detailed search logic and path list provided in references/sapi-endpoints.md.
- [COMMAND_EXECUTION]: The skill relies on shell commands (curl, openssl, grep, sed) to interact with APIs, generate signatures, and extract credentials from local files.
- Evidence: Implementation examples and bash scripts in references/authentication.md and references/sapi-endpoints.md.
- [PROMPT_INJECTION]: The skill exhibits surface area for indirect prompt injection by processing external data from Binance public and private APIs without explicit boundary markers or sanitization of returned content before presentation.
- Evidence:
- Ingestion points: API responses from get_capabilities, get_buy_and_sell_payment_methods, and authenticated order history.
- Boundary markers: Absent.
- Capability inventory: Shell execution for curl, openssl, and file system read access.
- Sanitization: Absent for data returned from external APIs before it is incorporated into the response.
Recommendations
- AI detected serious security threats
Audit Metadata