binance

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill's installation process downloads and executes a shell script directly from the author's official GitHub repository (binance/binance-cli) using a 'curl | sh' pattern in the SKILL.md frontmatter.
  • [EXTERNAL_DOWNLOADS]: Fetches the binance-cli installer and binary releases from the official Binance GitHub organization, which is the expected source for this vendor skill.
  • [COMMAND_EXECUTION]: The skill relies on the execution of the binance-cli binary to interact with the Binance API. It translates natural language requests into complex CLI commands with various parameters for spot, futures, and margin trading.
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates sensitive financial operations based on user input.
  • Ingestion points: The agent ingests user-provided arguments for command parameters (e.g., symbols, amounts, order types) which are then used to construct binance-cli commands.
  • Boundary markers: No explicit boundary markers or 'ignore embedded instructions' warnings are present in the reference files to delimit untrusted data.
  • Capability inventory: The skill possesses high-impact capabilities including order placement (new-order), asset transfers (universal-transfer), and withdrawals (withdraw) via shell command execution.
  • Sanitization: The skill instructs the agent to require the user to type 'CONFIRM' before executing production transactions, providing a manual authorization step that mitigates accidental or malicious execution of sensitive commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 08:07 PM
Security Audit — agent-trust-hub — binance